The Evolution of Religion and Morality project: reflections and looking ahead

Abstract

This special issue marks the formal end to the Evolution of Religion and Morality project and highlights the overall findings with particular attention to our second wave of data collection. In this concluding article, we first briefly detail how the project came about and how it developed. We then catalogue our contributions, summarizing the empirical results of key synthetic investigations that were part of the overall project. In an effort to shed some light on issues future researchers might benefit from knowing about, we also discuss some of the limitations and problems in design and execution of our effort. We conclude with a discussion of current, ongoing works, and our vision for the future of the cognitive and evolutionary studies of religion.

Full text

Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=rrbb20
Religion, Brain & Behavior
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/rrbb20
The Evolution of Religion and Morality project:
reflections and looking ahead
Benjamin Grant Purzycki, Martin Lang, Joseph Henrich & Ara Norenzayan
To cite this article: Benjamin Grant Purzycki, Martin Lang, Joseph Henrich & Ara Norenzayan
(2022) The Evolution of Religion and Morality project: reflections and looking ahead, Religion, Brain
& Behavior, 12:1-2, 190-211, DOI: 10.1080/2153599X.2021.2021546
To link to this article: https://doi.org/10.1080/2153599X.2021.2021546
Published online: 06 Apr 2022.
Submit your article to this journal
Article views: 235
View related articles
View Crossmark data

TARGET ARTICLE
The Evolution of Religion and Morality project: reflections and
looking ahead
Benjamin Grant Purzycki
a
, Martin Lang
b
, Joseph Henrich
c
, and Ara Norenzayan
d
a
Department of the Study of Religion, Aarhus University, Aarhus, Denmark;
b
LEVYNA Laboratory for the
Experimental Research of Religion, Department for the Study of Religions, Masaryk University, Brno, Czechia;
c
Department of Human Evolutionary Biology, Harvard University, Cambridge, MA, USA;
d
Department of
Psychology, University of British Columbia, Vancouver, Canada
ABSTRACT
This special issue marks the formal end to the Evolution of Religion and
Morality project and highlights the overall findings with particular
attention to our second wave of data collection. In this concluding
article, we first briefly detail how the project came about and how it
developed. We then catalogue our contributions, summarizing the
empirical results of key synthetic investigations that were part of the
overall project. In an effort to shed some light on issues future
researchers might benefit from knowing about, we also discuss some of
the limitations and problems in design and execution of our effort. We
conclude with a discussion of current, ongoing works, and our vision for
the future of the cognitive and evolutionary studies of religion.
ARTICLE HISTORY
Received 12 October 2020
Accepted 20 April 2021
KEYWORDS
Evolution; religion; morality;
cooperation; field research
1. Introduction
After a decade of effort, the Evolution of Religion and Morality project (ERM project henceforth) is
coming to a close. In this paper, we take a look at the early motivations of the project and highlight
some of the practical issues involved in organizing and executing a project like ours. In the spirit of
self-assessments found in the discussion sections of standard scientific reports, this article discusses
some noteworthy features and limitations of an entire project. We first briefly detail the genesis of
the project, followed by a discussion of what it takes to organize cross-cultural field research, fol-
lowed by a discussion of some of the limitations of our analytical and methodological execution. We
hope that others will find the discussion useful if they choose to embark on similar ventures.
2. Genesis and theoretical motivation
The Evolution of Religion and Morality project was the result of initial collaboration between Ara
Norenzayan, Ted Slingerland, Joe Henrich and Mark Collard. These conversations grew first into an
informal umbrella organization called the Center for Human Evolution, Cognition and Culture
(HECC) that included a subproject on religion that aimed to assemble a network of researchers,
which they dubbed the Cultural Evolution of Religion Consortium (CERC
1
). After a failed attempt,
the group landed a nearly $3 million Project Grant in 2012 to study the evolution of religion and
morality using a combination of ethnographic and experimental methods (the experimental wing
headed by Henrich and Norenzayan) and designing a large-scale historical database (the database
wing led by Collard and Slingerland). Funding from the John Templeton foundation later
© 2022 Informa UK Limited, trading as Taylor & Francis Group
CONTACT Benjamin Grant Purzycki bgpurzycki@cas.au.dk
RELIGION, BRAIN & BEHAVIOR
2022, VOL. 12, NOS. 1–2, 190–211
https://doi.org/10.1080/2153599X.2021.2021546

supplemented the ethnographic-experimental wing of the project. Adam Barnett joined the project
to effectively formalize the organization, along with four post-docs, two of whom –Benjamin Pur-
zycki and Martin Lang –led the experimental part of the project that comprised two waves of data
collection across 15 field sites.
The CERC project cumulatively built upon several prior collaborative projects, including the
Culture and the Mind Project
2
led by Steve Laurence and Steve Stich (Barrett et al., 2016), the Vir-
tues in Conflict Project led by Dan Hruschka and Charles Efferson (Hruschka et al., 2014) and the
Roots of Human Sociality
3
Project led by Jean Ensminger and Joe Henrich. Specifically, we used a
random allocation game across both waves came from the Virtues in Conflict Project and an early
iteration of our version of the game was administered in Fiji (McNamara et al., 2016). An initial
empirical insight came from post-hoc analyses in the Roots of Human Sociality Project (Ensminger
& Henrich, 2014; Henrich et al., 2010). As we did in the ERM project, these precedent studies all
experimentally investigated cross-cultural patterns in cooperation. However, the cognitive and
evolutionary sciences of religion had not really engaged in large-scale cross-cultural studies, com-
bining several methodological approaches, in such an organized fashion before. We sought to
change that.
In fact, there were several interrelated aims that motivated the project. One was the creation of an
international collaborative network of researchers to study the evolution of religion. Like the rest of
the academy, every scholar who studies religion is trained in a specific home discipline, with its own
set of perspectives, research tools, and, inevitably, blindspots. Discipline-centered research leads to, at
best, a partial understanding of complex and multi-layered sociological phenomena like religion. The
complementary expertise from different disciplines (e.g., anthropology, history, psychology, religious
studies), and the deployment of studies across a range of populations, was designed to facilitate high-
quality, high-impact research that cuts across disciplinary boundaries and synthesizes a wide range of
knowledge that otherwise would be prone to fragmentation. We hoped that this collaborative
environment would foster dialogue across disciplinary divides, including science-humanities inte-
gration (Slingerland & Collard, 2011) and stimulate further advances in the evolutionary study of reli-
gion. A second aim was to create opportunities for a new generation of promising graduate students,
postdocs, and other early-career scholars to hone their research skills and deepen their understanding
of religion in this interdisciplinary framework. Our third aim was to disseminate findings from this
project as broadly as possible—to scientific audiences, to scholars in the humanities, as well as to the
general educated public to enrich understanding of religion’sroleinsociety.
The core theme of the CERC project was to examine how cultural evolutionary processes have
shaped aspects of religions over time and across populations, and how those aspects in turn have
shaped human sociality, throughout cultural history and in the contemporary world. The key theor-
etical question was whether modern world religions are products of a long-term cultural evolution-
ary process—driven by both competition and conflict among cultural groups and shaped by reliably
developing features of human cognition—that has gradually assembled and integrated combi-
nations of supernatural beliefs, rituals, devotions, and institutions that instill commitment and pro-
mote large-scale cooperation. Integrating a growing set of theoretical ideas from the cognitive and
evolutionary sciences (Atran & Norenzayan, 2004; Boyer, 2001; Bulbulia et al., 2008; Bulbulia &
Sosis, 2011; Henrich, 2004; Johnson, 2005; McCauley, 2011; Sosis & Alcorta, 2003; Wilson,
2010), the project aimed to harness the world’s religious diversity to answer a set of interrelated
questions: Are some religious beliefs, rituals, and credible displays of faith linked to within-group soli-
darity and cooperation? If so, how far does this solidarity and cooperation extend? What are the par-
ticular features of supernatural agents that might have coevolved with human prosociality in a
cultural evolutionary process of ever larger and more complex societies? To what particular domains
of behavior did the concerns of moralizing gods apply and how cross culturally similar or different are
these?
4
We broke these guiding questions down into a set of related sub-questions that stretched from
individual-level variation in cognitive processes to behavioral and economic measures all with
RELIGION, BRAIN & BEHAVIOR 191

implications for societal-level social dynamics and deep historical patterns. Of course, these ques-
tions had been seriously pondered and addressed in many different ways since before the dawn of
social science. But by building on this legacy and directly applying state of the art methods and cur-
rent theories of cultural evolution and the cognitive science of religion, we hoped to offer a few key
contributions that would both raise the methodological bar and propel ongoing inquiry while
remaining mindful that all research projects have their limitations.
3. Contributions
The aim of the ERM project was to combine the testing of top-down theories with exploration from
the bottom-up. As part of the bottom-up effort, the project aimed to foster a rich ethnographic
component–which influenced the choice of postdocs. We felt this was particularly important
given the lack of a collaborative ethnographic science of religion, certainly not one that employed
comparable, quantitative ethnographic methods. Protocol development for the first wave of the
project took shape after our first team meeting in the spring of 2013. We drew from what we learned
and expanded the project for a second wave which took place primarily in the summer of 2015.
Across these two waves, our team collected behavioral and survey data from over 2,300 partici-
pants across 15 field sites. After implementing a host of error-detection procedures (e.g., auditing
data entry and data-checking), our final data set comprised 2,228 participants who varied in their
socio-economic status, means of subsistence, household size, and, importantly, the type and inten-
sity of religious beliefs and behaviors. We strove to sample from diverse populations to address the
ongoing challenge of narrow sampling in the rapidly growing fields of the psychology of religion
and the cognitive science of religion (Boas, 1930; Henrich et al., 2010; Kressel, 1990; Norenzayan,
2016; Saroglou & Cohn, 2013; Sears, 1986).
Over two waves, the project has directly produced over 20 empirical reports, one article detailing
the data set and methods from the first wave (Purzycki et al., 2016a), and the present overview
paper. The project indirectly went toward even more empirical reports, two dissertations (Baimel,
2019; Vardy, 2019), an R package designed to process ethnographic free-list data and perform cul-
tural consensus analysis (Purzycki & Jamieson-Lane, 2017), three pre-registrations (including the
second wave data analyses
5
), and an open data set comprising over 2,000 individuals.
In terms of our contribution to the field of cognitive and evolutionary science of religion, and
our understanding of the evolution of religion and morality specifically, the project delivered sev-
eral important findings. First, surveying the religious landscape across our sites confirmed that
people indeed hold diverse beliefs about the abilities and interests of their gods. At most sites,
we asked a systematic battery of questions about (including beliefs and ritual devotion frequency
toward
6
) two gods: one god that was deemed to be relatively more interested in human interperso-
nal conduct who punishes people for immoral behavior and has a broad range of knowledge (a
“moralizing god”) and the other varied in these qualities (we refer to the latter gods as “local
gods”simply for the sake of distinction). We selected these gods on the basis of preliminary ethno-
graphic interviews, and, in most sites, all individuals were asked about the same two gods (some
participants in two of the samples answered questions about a different local god). With respect
to gods’reported punishment, rewarding abilities, and their knowledge breadth (i.e., omniscience),
our data revealed substantial between-site variation; but crucially, reported belief in punishment
and monitoring abilities of moralizing gods was on average higher than that of local gods, and,
at most sites, the maximum punishment-monitoring score for moralizing gods was the mode of
the distribution. This validated our design and selection process (see note 6). See Figure 1 for a
graphical overview and Supplementary Material in Lang et al. (2019) for a detailed overview of
assessed gods’qualities and their between-site variation.
In line with previous studies (Purzycki, 2011,2013; Purzycki et al., present volume), our data also
showed that classifying gods as “local”does not necessarily imply a lack of punishment-monitoring
abilities or moral concern. Consistent with our design of selecting deities based on preliminary
192 B. G. PURZYCKI ET AL.

ethnographic work, we confirmed that such gods are often relatively less interested in interpersonal
conduct and are believed to have relatively less punishment and monitoring abilities compared to
moralizing gods, indicating that moralization, punishment, and knowledge people ascribe to super-
natural agents is a continuum, not a strict dichotomy. Further supporting this point, the detailed
analyses conducted by Purzycki et al. in this special issue demonstrate that the probability that
people will report at least some moralistic punishment-monitoring abilities of their local gods
when explicitly asked is higher than chance, even after holding constant any correlation with mor-
alistic gods (cf. Placek & Lightner; Vardy & Atkinson, present volume). This may be partly due to
individual-level biases in human cognition that naturally recognize agents as having minds with
moral interests (e.g., Gray et al., 2012; Krátký et al., 2016).
The next big question the current project aimed to tackle was identifying factors that could
account for differential commitment to moralizing gods. Previous theories have suggested that
Figure 1. Density plots displaying the distributions of belief in punishment and monitoring abilities of moralizing (full line) and
local (dashed line) gods.
RELIGION, BRAIN & BEHAVIOR 193

one of the most important factors predicting commitment to supernatural agents is existential inse-
curity. Among the most prominent theories, two competing hypotheses proposed are: (1) existential
insecurity motivates people to engage in religious beliefs and behaviors because powerful and pro-
tective supernatural agents provide a sense of safety and control in a threatening environment (Nor-
ris & Inglehart, 2011; Henrich et al., 2019); and (2) existential security allows individuals to delay
immediate benefits and explore new mechanisms such as belief in moralizing gods to govern larger
societal formations (Baumard & Chevallier, 2015). The data lend support to the former hypothesis,
showing that material insecurity (insecurity about food availability in the future) positively predicts
mental commitment to moralizing gods.
Interestingly, however, this relationship is reversed for local gods (Figure 2A summarizes the
results found in Baimel et al. in this special issue). Addressing the existential security hypothesis,
Purzycki, Ross, et al. (2018) provide detailed analyses and additional discussion suggesting this
prediction lacks support at the level of individuals. Moreover, Weigel’s report in this special
issue includes an analysis of material insecurity and positive contributions to distant co-religio-
nists and outgroups, which may function as safety nets for materially insecure people. The com-
bination of these two results suggests that socioecological shocks may simultaneously increase
faith in moralizing gods while diminishing commitment to local gods in some contexts (McNa-
mara & Henrich, 2018).
Another factor predicting commitment to moralizing gods that the current project explored was
participants’gender. It has long been suggested that women report higher commitment to super-
natural agents compared to men (Argyle & Beit-Hallahmi, 1975; Stark, 2002), although this suppo-
sition was usually assessed only using samples from Western countries, thus limiting the scope of
hypothesis-testing. Furthermore, explanations for this gender gap have been inconclusive due to
conflicting evidence regarding underlying mechanisms. Using the sample from the ERM project,
Vardy et al. (present volume) show that our data from 14 small-scale societies indeed indicate
the presence of a gender gap in commitment to moralizing gods –but not local gods. Testing var-
ious theories regarding potential mediating factors for the gender gap, years of formal education
appear to be the strongest mediating factor such that women on average have fewer years of formal
education yet report higher commitment to moralizing gods. This result suggests that access to for-
mal education acts as a “secularization force”that exposes individuals to values outside of their reli-
gious tradition (Berger, 1967) and, because men have better access, they are more secular. Figure 2B
suggests this relationship by plotting z-scored values for years of formal education and commitment
to moralizing gods by sex. The figure shows that while women on average score higher on commit-
ment to moralizing gods compared to men, their average length of formal education is shorter (see
Vardy et al., present volume for a rigorous mediation analysis).
The central goal of the ERM project was to test whether different supernatural concepts and
differential degrees of commitment to punitive and omniscient gods affect prosocial behavior in
economic games. Across the 15 sites, our team deployed a random allocation game while
second-wave participants also played dictator games. In these games, participants individually allo-
cate endowed money to various beneficiaries. In four rounds of either game, participants allocated
coins between themselves, local anonymous co-religionists, geographically distant anonymous co-
religionists, and geographically distant anonymous outgroups. Our main results, detailed in several
publications (Lang et al., 2019; Purzycki et al., 2016b; Purzycki, Henrich, et al., 2018), showed that
rating moralizing gods as punitive and omniscient positively predicts allocations to distant co-reli-
gionists when participants allocated money between the distant co-religionist cups versus self and
local co-religionists cups. In the random allocation game, rating moralizing gods as punitive and
omniscient increased the odds for a coin landing in the distant co-religionist cup by 26%, while
in the dictator game, this belief was associated with one more coin in the distant co-religionist
cup (an increase of 36% relative to rating gods as not punitive and omniscient). Figures 2C-D illus-
trate the effects for beliefs in the self vs. distant co-religionist games. Together with a host of control
variables used to adjust for this effect and the fact that we did not observe similar effects for rating
194 B. G. PURZYCKI ET AL.

moralizing gods as rewarding or for the punishment-monitoring ratings of local gods, these results
suggest (with caveats, see below) that belief in moralizing gods plays a role in facilitating
cooperation between anonymous co-religionists.
The results from games that pitted distant co-religionists, outgroups, and participants against
each other were more equivocal. While rating moralizing gods as punitive and monitoring generally
predicted preferring a co-religionist before an outgroup member, these estimates were uncertain
and varied across sites. Similarly, we detected a noisy signal in the self vs. outgroup games where
the punishment-monitoring scores predicted positive allocations to outgroups; however, this
Figure 2A -D.#A simplified overview of the main findings of the Evolution of Religion and Morality Project. (A) Material insecurity
increases the commitment to moralizing gods but acts in the opposite direction regarding the commitment to local gods; (B)
Women report higher commitment to moralizing gods than men do and differential access to formal education partially accounts
for this difference; (C) The effects of rating moralizing gods as punitive and omniscient on odds of allocating a coin to distant/
outgroup cups in the Self vs. Distant co-religionist and Self vs. Outgroup rounds of the random allocation game; (D) The effects of
rating moralizing gods as punitive and omniscient on allocations to distant/outgroup cups in the Self vs. Distant co-religionist
and Self vs. Outgroup rounds of the DG. Note that for the C and D plots, the Y-axes were truncated at impartial allocations (odds
of 1 in random allocation game and allocations of 5 in dictator game) to highlight the detected differences. All plots simplify the
complexity of our results and are only illustrative. For more details and discussion, see Baimel et al., present volume for panel A,
Vardy et al. (present volume) for plot B, and Lang et al. (2019) for plots C and D.
RELIGION, BRAIN & BEHAVIOR 195

effect seemed to be driven by avoiding hyper-unfair offers in the dictator game as well as by Chris-
tian sites (see Lang et al., 2019 and Figures 2C-D).
Finally, the project also attempted to tackle the causal relationship between beliefs in moralizing
gods and prosociality by using various types of control primes and primes with moralizing gods,
local gods, and secular authority. Results did not provide clear support for the causal connection
between the reminders of moralizing gods and prosocial action. Specifically, we did not observe
substantial differences between the moralizing god and control primes, although primes with
local gods and secular authority appeared to promote selfish and parochial favoritism in some
places. While between-site variability in the effects of primes and their unclear ecological validity
undermines the strength of this method, we believe that the lack of difference between the control
and moralizing god primes suggests that even without primes people mostly played fairly and the
primes had only low variation to act upon in the first place. It might have been the case that people
were inadvertently primed in all of our experiments by the cup labels that often identified recipients
as believers in the specified moralizing gods. In other words, the actual moralizing god primes
might have been redundant. Unfortunately, we have no comparison available to draw clear infer-
ences (see below for more discussion on our design). On the other hand, the parochial bias of local
gods suggests an interesting venue for future research (see Soler, Purzycki, and Lang, present
volume) as does the more recently developed supernatural framing procedure that finds that remin-
ders of a moralizing God and of Karma reliably increases generosity in the dictator game among
Christians, Hindus, and Buddhists (White et al., 2019).
4. Moving forward: lessons learned and confirmed
We now turn to potential pitfalls and lessons learned that future projects should avoid. Indeed,
there are many features particular to large-scale projects that researchers should anticipate. Any
project has its problems and limitations, and one of this size and scope inevitably suffers from
sources of error. Ours was no exception. This section addresses some of the practical, theoretical,
methodological, and analytical issues we identified, and makes some recommendations for areas
that might deserve more attention.
4.1. Working with people
4.1.1. The team
By far, the single most important ingredient for a successful cross-cultural project like ours is the
recruitment of, and collaboration with, a diverse set of engaged researchers who are willing to col-
lect data in a field setting (we were fortunate in this respect). However, assembling the right team
can be challenging for several reasons. First, even if one can find potential partners who concur with
one’s scientific epistemology, disciplines like anthropology and religious studies, which provide cru-
cial insights into particular religious phenomena, still lack a tradition of team collaborations. The
professional payoffto such efforts for researchers hailing from those disciplines is often unclear.
These disciplines also cultivate an independent spirit that does not prioritize adhering closely to
tight protocols and procedures when local context would seem to favor a deviation.
Second, an ideal team would have consisted entirely of long-term ethnographers, fluent in the
local language and context, trained in quantitative methods, and focused on the detailed investi-
gation of religion, myth, magic, and ritual in their respective field sites. Since we drew most of the
researchers for our team from Henrich, Norenzayan, and Slingerland’s network, the members of
our team had variable expertise in the study of religion. To minimize the same kind of variability,
future research can instead make an open call for researchers and/or specifically recruit research-
ers based in field sites with variation important to the study subject and provide training in
methods where needed. Given some of the recent developments in the field of the cognitive
196 B. G. PURZYCKI ET AL.

and evolutionary science of religion (see Geertz, 2020), such a recruitment method should be
increasingly easier to attempt.
Finally, if possible, we recommend that projects of this size should have two managing post-docs
throughout their duration, especially if they want to go to the field. Managing a project of this size
and scope can be fairly consuming work; in addition to keeping on top of data processing, compil-
ing, cleaning, and analyzing, project management includes team communications, aligning, facili-
tating, and accommodating the various skill sets of team members, general quality control and
problem-solving, and coordinating work on academic deliverables. Overall, management of a pro-
ject of this size requires coordinating multiple tasks that overlap in time and may extend well
beyond the project’s funding time frame. More workforce available at the same time would help
alleviate the burden from a single post-doc. Keep in mind, too, that smooth transitions between
managers also requires substantial effort; it is prudent to anticipate that any incoming postdocs
will need some time to acquaint themselves with the project. As is the case generally, good com-
munication, documentation, and organization will facilitate smoother transitions.
4.1.2. The field, site selection, and application
The field sites from which we sampled (Table 1) were selected indirectly through the research team;
we selected field researchers more than the field sites. While this might have had the benefit of expe-
diting the work, this decision also had some consequences. For example, one byproduct of our
selection of researchers was that Christianized populations were over-represented in our sample.
While we had two Hindu samples and one Buddhist site, we had no Muslim groups. This gave
us little breadth to explore, among other things, how much proselytizing religions have interacted
with local traditions. Does the historical depth of Christianity present at sites, missionary presence,
and/or geographic proximity to missions affect individuals’behavior (Woodbury, 2012)? The type
of data we collected did not have a sufficient group-level sample size to reliably estimate the degree
to which local traditions have influenced and been influenced by universalizing traditions.
Moreover, as fieldwork is challenging, researchers may fail to collect the required data because a
new political appointee suddenly pulled their research visa, a civil war started, a hurricane hit, or
they got malaria, just to name a few (see BorgerhoffMulder et al., 1996; Hewlett, 2019). The lesson
here is to choose members wisely, establish working groups to discuss research norms for altering
protocols, and prepare for a degree of attrition—some members of the team will not be able to deli-
ver for some unforeseeable reason.
The biggest challenge to implementing our design revolved around ensuring that field research-
ers are actually doing the same thing in each site. If there was any unexpected variation that needed
to be accounted for in situ, some researchers made decisions to collect additional data (or simply
not collect particular variables). Others used different scales to accommodate variation, but lacked
the time to ensure that they had construct validity (see below for more general issues). Others made
mid-stream decisions about which gods to use in the study, thus splitting the sample along a crucial
dimension of the study. In a similar way, there was some variation in how we recruited participants
across sites. Aside from the inevitable contextual differences (e.g., sampling entire villages versus
door-to-door sampling in urban areas), some samples are demographically biased; some, for
example, are all relatively older while others included mostly women.
To address these issues statistically, we held demographics constant in our regression models
and treated our field sites as simple (i.e., “fixed”)effects to study only the individual differences
within sites (see Gelman & Hill, 2007, pp. 251–278; Silk et al., 2020 for recent critical examination
of this and other challenges in modeling multilevel data). Among other issues, however, this
method does not account for the fact that participants from different sites might systematically
answer scales differently. For example, in measuring ritual frequency, some very religious samples
(e.g., Yasawan Christians) might only answer at the level of weekly devotions (e.g., going to Church)
while others might perform small daily devotionals, thus making them look “more religious”. While
centering on site-specific means (i.e., subtracting individuals’data from the site-specific average)
RELIGION, BRAIN & BEHAVIOR 197

Table 1. Details about participants, games, gods, and allocation recipients. Prime column indicates which experimental conditions were used in field site (M –moralistic god; L –local god; S –secular
prime; C –control prime).
SITE COUNTRY N PARTICIPANT LOCAL DISTANT OUTGROUP PRIME
MORALIZING
GOD LOCAL GOD
CACHOEIRA Brazil 274 A member of
Candomblé in
Cachoeira
Another Candomblé
member in Cachoeira
Candomblé member
in Salvador
Evangelical living in
Salvador
MLSC Christian God Candomblé God
(Ogum)
COASTAL TANNA Vanuatu 178 A person from
Christian village
Another person from the
same village
Christian from a
different village
Another person from a
Kastom village
MLC Christian God Garden spirit
(Tupunus)
HADZA Tanzania 201 A person from Hadza
camp
Hadza living in the same
camp
Hadza living in
another camp
Datoga living in another
camp
–Haine
(Traditional)
Ishoka
HUATASANI Peru 94 A person from
Huatasani or a
nearby village
Catholic from the same
town/village
Catholic from
Quechua Taraco/
Aymara Vilque
Chico
d
Evangelical from Quechua
Taraco/Aymara Vilque
Chico
MC Christian God Mountain Spirits
INLAND TANNA Vanuatu 112 A person from Kastom
village
Another person from the
same village
Another person from a
Kastom village
Another person from a
Christian village
MC Kalpapan
(Traditional)
Garden spirit
(Tupunus)
KANANGA DRC 200 A born-again Christian
of Luluwa ethnicity
in Kananga
a
Another born-again
Christina of Luluwa
ethnicity in Kananga
Another born-again
Christian in Kananga
but not Luluwa
Not member of born-
again Christian and not
Luluwa but living in
Kananga
MLSC Christian God Kadim/Ancestor
spirits
LOVU Fiji 76 A Hindu from Lovu Hindu in Lovu Hindu from another
island in Fiji
–MC Hindu Bhagwan None available
MARAJó Brazil 77 A person from Marajó
b
Evan./Cathol. from Marajó Evan./Catholic from
Rondon
–MC Christian God Virgin Mary
INDO-MAURITIANS Mauritius 245 A Hindu from from La
Gaulette
c
Another Hindu from La
Gaulette
Hindu from Point aux
Piments
Muslim from Mahebourg MSC Hindu Shiva Ghost (Nam)
MYSORE India 165 A Hindu from Mysore Person from the same
neighborhood
Hindu Tamil Nadu Christian from Tamil Nadu MC Hindu Shiva Chamundeshwari
SAMBURU Kenya 40 A person from
Mbaringon
Another person from
Mbaringon
Christian Samburu
from Poro
Samburu from Poro –Christian/
Traditional
(Nkai)
None available
SURSURUNGA Papua New
Guinea
163 A person from Nokon Person from your clan Person from your
church
Stranger from far away MLC Christ. God
(Káláu)
Spirit (Sírmát)
TURKANA Kenya 247 A person from
Nadome
Turkana who attends
church in Lobangaa
(closest church to
Nadome)
Turkana who lives and
attends church in
Awar Naparan
Turkana who lives in Awar
Naparan but nevr
attends church
MLSC Christ. God
(Akuj)
Ancestor spirits
TYVANS Tyva
Republic
81 A Buddhist from Kyzyl Buddhist from Kyzyl Buddhist from Ak
Dovurak
–MC Buddha Burgan Spirit-masters (Cher
eezi)
198 B. G. PURZYCKI ET AL.

YASAWA Fiji 75 A person from Yasawa A person from “your
village”
person from another
island who goes to
church
–MC Christian God Ancestor spirits
(Kalou-vu)
a
Participants were asked about their religious identity in another large screening survey and were subsequently randomly sampled from this larger pool.
b
Participants were subsequently identified as either Catholic or Evangelical and this fed into other players’identities.
c
Participants were recruited from Hindu communities and not asked about their religion during recruitment.
d
Participants may have been sensitive to the language spoken in the distant village (Quechua or Aymara). Moya therefore used Taraco for half of the sample as the distant town and Vilque Chico for
the other half.
RELIGION, BRAIN & BEHAVIOR 199

might help in cases where there is enough variance, knowing more about the field sites themselves
can facilitate important qualifications of data weighting and interpretation. Establishing measure-
ment invariance of used scales as well as of behavioral variables across sites is another important, yet
challenging task for future studies (Fischer & Poortinga, 2018).
Regarding the implementation of experimental designs across different field sites, we recommend
that researchers both develop a detailed protocol and then thoroughly train researchers in implemen-
tation. Specifically, the implementation process might start by piloting efforts at one field site where
researchers video record a near-final protocol with a voice-over explaining the entire process, step-by-
step. Then, sitting together as a group (virtually if necessary), the team should watch the video
together with a draft protocol in front of them, stopping the video to ask questions and clarify subtle-
ties, thereby further detailing the protocol and clarifying any uncertainties. Then, at each field site, the
researchers could video record a pilot run of the protocol and send this and their datasheet back to the
lead researchers for a check (if possible) before proceeding. This is especially important if the
researcher is inexperienced with either the field site or experimental protocols. The lead researchers
can review this and provide feedback or deal with any site-specific challenges. If the field site remains
offthe grid, the researcher might video record the way they conducted experiments for later review.
Alternatively, if time permits and researchers are inclined to (these conditions do not always hold),
each researcher might pilot their protocol during a first field season, review and perfect it with the
team leads over the intervening year, and then deploy the final version in a second field season.
Finally, meta-data of any remaining or unforeseen deviations can substantially aid the leading analyst
of the project in clarifying irregularities in the observed results.
4.2. Working with theory and data
4.2.1. Focal contrasts, design, and interpretation
As noted already, our project lends itself to testing a wide range of hypotheses. We can formulate
our primary motivating predictions as four key contrasts. First, we used the prime conditions to
assess the causal effect that the motivations underlying moralistic, punitive, and knowledgeable
gods have on prosociality. Second, our primary published correlational results emphasize the
association between the expressed beliefs about these gods and prosociality, while a third contrast
focuses on how broad the effect of this association is (i.e., between self, local co-religionists, distant
co-religionists, and outgroups). However, a few interrelated issues suggest some modesty and cau-
tion with interpreting various contrasts involved in our studies is warranted.
Recall that most of the recipients of the coin allocations were co-religionists (local and distant).
Cups were generally assigned to participants’co-religionists of the moralistic god type (Table 1),
gods of which we selected based on preliminary ethnographic interviews. In other words, either
by default (e.g., everyone in a village identifies as Christian as in the case of Yasawa) or by design
(e.g., Buddhist Tyvans were recruited), participant affiliation with the moralistic god tradition was
an important aspect of participation generally.
7
If participants’religious affiliation –or any other
variable unaccounted for–contributed both to beliefs about moralistic gods as well as willingness
to behave prosocially towards co-religionists, target estimates may be biased. In other words, if
an unmeasured factor that varies cross-culturally has an influence on both reported moralistic
god beliefs and prosociality toward co-religionists, we provided nothing in the way of a mediator
to block any indirect effects that run through beliefs. Even simple variables such as material security,
age, sex, or more complicated factors like social desirability of adherence to moralistic god tra-
ditions, and so forth could have backdoor effects on game outcomes and including them all as
simple effects in regressions can give a misleading impression of target effects without also includ-
ing, for example, a mediator between beliefs and outcome (see Pearl et al., 2016 for this and other
problems with confounds and causal inference). This too stresses the importance of carefully think-
ing through a causal model with resulting cross-sectional or longitudinal data in order to appropri-
ately condition the predicted effects on as few but necessary variables as possible.
200 B. G. PURZYCKI ET AL.

Furthermore, a fourth contrast, namely, the prediction that moralistic deity traditions are more
likely to be associated with extended prosociality than the local traditions, was never adequately
established largely because recipients (including players) were associated with the moralistic
deity traditions (cf. Boyer, 2020). The set-up for economic games assumed an endowment division
between co-religionists. We defined co-religionists based on a shared moralistic god, not the local
god (except for one site; see Soler et al., this issue). In other words, our primary reported contrast was
not between tradition types –that is, between moralistic traditions and nonmoralistic traditions –but
instead between the kinds of beliefs that moralistic god constituents have about their moralistic deities.
While adjusting our models for the effects of belief in local gods is crucial to show that not just any
belief in supernatural agency promotes extended prosociality toward co-religionists of the moralis-
tic deity tradition, a proper test of the contrast between traditions would entail comparing allo-
cations between recipients identified with the local god tradition and recipients of the moralistic
god tradition.
8
4.2.2. Statistical modeling strategy
Our theory generated several causal predictions regarding the effects of belief in moralistic gods on
extended prosociality. Although manipulating actual religious faith is not feasible, we aimed to
probe the causal relationship by using a previously established method of priming participants
with subtle reminders of various deities and secular authorities. Priming people with indices of reli-
gion, we reasoned, would harness the effects of beliefs, but as we discussed above, the results were
unclear at best (see Nichols et al., 2020; White et al., 2019; and Pasek et al., 2020, for more robust
causal evidence and recent research developments that were partly inspired by this project).
Because our primes did not systematically work in the same way across sites, our main published
results and planned contrasts relied on correlational data analysis. That is, we used self-reported
beliefs about moralistic gods to predict behavior in economic games, which supplied crucial indi-
cators for the proposed causal relationship. As noted above, in practical and absolute terms, the
effect was small and it is open to question whether this effect approximates the extent of cooperation
necessary to sustain social complexity, as predicted by the theory. Moreover, correlational analyses
further limit the kinds of inferences we can draw. This section raises some of these limitations and
proposes possible solutions.
First, although our theoretical understanding of the underlying cultural evolutionary processes
draws on a broad range of formal mathematical models including models specifically built to help
us think about the interaction of rituals, supernatural belief and costly action such as cooperation
(e.g., Henrich, 2009; Wildman & Sosis, 2011), our specificstatistical model does not correspond
directly to any specific formalization and ritual behaviors played virtually no role in our project.
Moreover, we based our only preregistration (for Wave II) on previous results and statistical mod-
eling strategies, not on any formal predictions that came from a precise mathematical model.
Second, to assess the stability of the effect of beliefs, we generated well over 100 total model spe-
cifications across our three main omnibus papers alone. At best, this approach, compounded by the
potential of adding countless “control”variables, was inefficient and appeals more to standard prac-
tices than pre-planned, modeled, and well thought-out contrasts. The standard approach entails a
“full model”that includes variables such as participant age, sex, and so forth as “standard”or
“necessary controls”(see Achen, 2005). Comparing this full but potentially confounded model
with a base model is assumed to check for the robustness of the target effect. While this might
be sufficient for assessing the presence of an effect, it does not help with selecting the most sensible
model (and model selection criteria tend to be biased towards confounded models), nor does it
indicate the reliability of an effect size.
Arguably, the standard approach to statistical model building and analytical strategies should
entail limiting the number of statistical models examined based on some principled theoretical
reasons drawn from a theoretical model designed to answer the specific question at hand. In hind-
sight, modeling and simulating first likely would have: (a) made predictions and focal contrasts
RELIGION, BRAIN & BEHAVIOR 201

clearer, (b) anchored analytical plans, (c) provided a tangible source of discussion in the greater aca-
demic community for feedback, and (d) highlighted blindspots or limitations of the theory’s specifi-
city (see Kandler & Powell, 2018; Wilson & Collins, 2019; and Gelman & Hill, 2007; pp. 155–166
and McElreath, 2020, pp. 49–70 for various hands-on examples and direct discussion of the
value of simulation and regression).
If we had simulated first, we could have preemptively explored, say, the relationships between
demographic variables and target variables (e.g., does a common unmeasured mechanism associ-
ated with sex influence beliefs and game outcome? Do we expect that, say, unmeasured group-
level factors might contribute to the outcome? Do we expect that the game outcome could cause
beliefs to shift?). Once more relatively grounded models are explored, subsequent simulations
could entail setting up the appropriate statistical models to assess predictions. Simulating the
other variables and the coefficient sizes would foster a sense of the world of possibilities that
exist analytically and the distance between theory and practices. Simulating makes questions
such as Does the theory predict a particular size of an effect? If so, what would it look like? How
does this effect correspond to the narrative of the theory? naturally emerge from this process, and
therefore feeds back into refining theory.
4.2.3. Missing and multi-level data
Large data sets are often peppered by random missing values, but also large swathes of data might be
missing for more systematic reasons. In our data set, some sites simply have no data for “local gods,”
whereas in other cases, some large sections of data are treated as missing due to complications with
scale intelligibility. In all of our main papers, we simply dropped these cases from consideration.
This is advisable in the case of random missingness because whatever process led to missing
data, likely had no effect on the experimental outcomes. However, if there is systematic missingness,
dropping cases is inadvisable. One option we generally avoided was imputation but there are useful
resources encouraging this practice (Azur et al., 2011; McElreath, 2020; van Buuren & Groothuis-
Oudshoorn, 2011).
While there are many imputation strategies, common approaches entail either replacing missing
values with data generated by some algorithm (e.g., a random selection of possible values) or by
entering some value presumed to be safe (e.g., entering the sample mean for all missing values
of a variable). Both fail to provide a principled simulation of the data generation process. The latter
will often provide misleadingly precise estimates when there is a lack of variation. While more com-
plicated than algorithmic approaches, one avenue that provides some promise stems from modeling
the missingness patterns and integrating that sub-model with the primary analytical model (see
Purzycki, Pisor, et al., 2018 for an example). This approach imputes missing data based on prior
probability distributions of the data type, thus making it both more principled and flexible than
the alternatives (see Gelman & Hill, 2007; pp. 529–543 and McElreath, 2020, pp. 499–516 for
further discussion).
9
As implied by the previous section, finding the best way to assess hypotheses with cross-cultural
data such as ours is a non-trivial task. For example, while we might expect the effect of supernatural
punishment to vary across sites, we did not have a clear or strong a priori hypothesis why it might.
We explored whether the supernatural punishment effects differ between Christian and non-Chris-
tian sites on the suggestion that religious proselytizing by Christian missionaries may positively
affect behavior toward outgroups. However, this exploration was not justified by any formal theor-
etical model and a proper test would require higher religious diversity in proselytization practices
(e.g., religious traditions such as Islam). In many model specifications, we treated field site as a
simple effect, while in others we model error with clustered robust standard errors, or modeled
them as superordinate categories in multi-level models. Again, this was for the purposes of robust-
ness checks more than driven by any principled motivation. In a follow-up paper (Purzycki, Pisor,
et al., 2018), we constructed a model fully varying the effects for gods’knowledge breadth and
202 B. G. PURZYCKI ET AL.

punishment across sites and found a stronger general relationship between omniscience and play-
ing fairly in the random allocation game than punishment and playing fairly.
In this section, we identified some intercorrelated issues with sampling, design, data treatment,
and analysis. While practical aspects of field research will inevitably complicate design (e.g., religion
is not about “groups”or there are no religious outgroups), most of these issues we addressed in this
section could have been avoided with explicit, formal models and simulations designed and
explored prior to execution. We now turn to the execution itself, with an assessment of our meth-
odological protocols and instruments.
4.3. Working with methods
4.3.1. Policies, detailed protocols, and data checking
One important lesson learned regards the management of large data sets, including the digitaliza-
tion of paper response sheets. To prevent data loss, we encouraged each researcher to scan their
paper sheets such that we would have available both the digital transcription of individual data
points as well as the original source material. This procedure allowed us to implement multiple
data quality checks, including auditing the quality of data transcription. We recruited graduate stu-
dents to help audit at least 30% of transcriptions at each site, controlling the accuracy of transcrip-
tion and marking potential problems. If the rate of found inconsistencies was alarming, we
encouraged researchers to enter the full data set once more, paying special attention to the proble-
matic variables. Our audits revealed several problems and deviations at each site and proved to be
indispensable in ensuring that our data were correct.
Apart from auditing data transcription, we also performed multiple data consistency checks
(e.g., checking if years of education or living in a city was less than participant age; seeing whether
or not the number of children participants had was feasible given their ages; ensuring that answers
to Likert scales were within the limits of the scale). These checks revealed further inconsistencies
and were either corrected according to the original paper sheet, or marked as NA if such a resol-
ution was not possible. In hindsight, while we performed many of these data quality checks in
the statistical software R, we performed the majority of data manipulation and cleaning in Micro-
soft Excel. While Excel offers a user-friendly visual interface, the fact that the history of these
manipulations is not directly recorded (and it is easy to make mistakes with large data sets) led
to many futile hours spent on comparing different versions of the data set and looking for the ori-
gins of discovered errors. For future research, we would strongly recommend performing all data
manipulation strictly in software where manipulation can be back-checked via statistical code and
the original files remain unchanged. Likewise, a very useful feature in this endeavor would be ver-
sion history as offered by software like GitHub, especially if multiple people are manipulating the
data set. At the end of the day, our workflows could certainly have benefitted from prior experience
and careful consideration early on (see Yenni et al., 2019).
4.3.2. Primes, games, and social complexity
As prime conditions were supposed to be culturally relevant and lack direct indices of agency (e.g.,
eyes), they were difficult (if not impossible) to standardize across contexts beyond these require-
ments. In other words, we should have more carefully ensured that the kinds of things our primes
were tapping into were consistent across sites (a Buddhist luck charm, for example, might have
primed people to keep more money, rather than prime Buddhist values of generosity; in another
context a Hindu icon might have been primarily associated with female commitments). More
recently, White et al. (2019; see also Pasek et al., 2020) developed a “supernatural framing”meth-
odology that could meaningfully be deployed across diverse populations. Participants are randomly
assigned to think about a particular culturally relevant deity or asked to take their deity’s perspec-
tive, just before their behavior is measured in an economic game or psychological task. Their behav-
ior is compared to a control group where this supernatural framing instruction is not given;
RELIGION, BRAIN & BEHAVIOR 203

alternatively, this procedure can be deployed in a repeated-measures design in which participants’
behavior is first measured without any instruction, followed by another measurement following
supernatural framing. In addition to the content of primes, we also employed different kinds of
primes ranging from contextual (e.g., a temple) and physical (e.g., a small statuette) to print
(e.g., images and text on tablecloths), which likely have different salience. This may further have
contributed to the between-site variation in detected effects of supernatural primes (Shariffet al.,
2016; cf. van Elk et al., 2015). Standardizing the medium for supernatural primes across sites
might yield results that are more reliable.
One question worth discussing is how satisfactory the dictator and random allocation games are
for measuring prosociality, especially the kind of broad cooperation typically appealed to by
researchers seeking to explain the development of social complexity. While our selected recipients
were explicitly associated with the kinds of cooperation people did not necessarily engage in
directly, and the random allocation game outcomes are correlated with market integration, the
question remains as to whether or not this would be an ecologically valid technique to assess the
kinds of expanded cooperation we see in socially complex societies. Supporting the argument
that behavioral games typically measure social norms for impersonal interactions (Ensminger &
Henrich, 2014), there is evidence from other studies that show that dice-rolling allocations like
the random allocation game are correlated with honesty in market transactions with strangers in
India (Kröll & Rustagi, n.d.) and behavioral game measures are correlated with voluntary blood
donations to strangers globally (Schulz et al., 2019), civic norms (Herrmann et al., 2008), voting
in U.S. elections (Fowler & Kam, 2007), and forest conservation in Ethiopia (Rustagi et al.,
2010); however, as has long been recognized (Gurven et al., 2008; Henrich et al., 2005), there is little
reason to expect behavioral games to measure dispositional social traits applicable across a broad
range of interpersonal, inter-ethnic, or familial relationships (see Pisor et al., 2020; Wiessner,
2009). Here, based on decades of research, explicitly theorizing what an experiment measures is cru-
cial (Henrich et al., 2010).
Another problem that complicated the design throughout the project was the fact that addres-
sing outgroups systematically was difficult. While the general theory rather straightforwardly
defines co-religionists as members of the same religious groups and outgroups as members of a
different religious group, finding such clearly defined groups at individual sites was not feasible.
The most challenging aspect of the design was to delineate clear grouping along religious lines
specifically from other sort of groupings, such as along ethnic or national lines. Moreover, religion
is not necessarily an affiliation in many places. For instance, there are not any obvious Haine believ-
ers who are not also Hadza, or Hadza individuals who would profess different religious affiliation
(apart from some Christian converts). Hence, outgroups had to be defined as ethnic outgroups (e.g.,
Apicella et al. chose the Datoga). Facing similar problems, other members of our team chose to
select outgroups as members of the same ethnic groups but professing no religious affiliation.
Yet at other sites, co-religionists and outgroups formed different denominations of the same reli-
gious traditions (Catholics vs. Evangelicals).
Finally, while we aimed to hold intergroup cooperation/competition/conflict constant between
these groups as these were assumed to crucially modify the effects of moralistic gods, we suspect that
our measures of emotional proximity or “identity fusion”were interpreted differently across sites
(Purzycki & Lang, 2019). At each site, we aimed to find groups with relatively neutral relationships
and further assessed each participant’s emotional closeness and perceived similarity to the selected
groups, aiming to account for the variance in inter-group relationships across sites in our statistical
models. These measures usually had unstable effects in our statistical models predicting behavioral
variation in the economic games, with the exception of the outgroup similarity and closeness
measures. These results suggest that there was likely a higher between-site variation in the relation-
ship to outgroups compared to a variation in the relationship to local or distant co-religionists.
Thus, future cross-cultural investigation of inter-group relationships need to take into account
that the hierarchical structure of multiple groups that an individual can belong to and their
204 B. G. PURZYCKI ET AL.

inter-group dynamic can substantially vary across sites and may significantly impact our under-
standing of general inter-group processes.
Indeed, it is no coincidence that one of the influential models of religious group survival (Sosis
& Bressler, 2003) assessed communes that are usually highly inclusive and easily separated from
other groups and thus approximate a better fit for theoretical models. However, investigating a
situation where the religious, ethnic, and national groupings simultaneously channel the flow
of information, resources, and competition is challenging. The fact that religious groups them-
selves –when available –are often part of various hierarchies and/or the subject of syncretism
between different religious traditions exacerbates this problem. Limited solutions may include,
for instance, using multidimensional scale design which may better reflect the difficulty of iden-
tifying appropriate outgroups and their hierarchical nesting. Another option might be to artifi-
cially create outgroups that could be easier to standardize, but this would entail trading off
realism and run the risk of being too inconsistently hypothetical for participants. Anticipating
such difficulties is important for future design.
5. Conclusion
Projects of this scope and scale quite often end with far more data than can be immediately used
without expanding the infrastructure. Given the labor required and how useful they are, many of
the elements of projects like this (e.g., protocols and datasets) should be thought of as “deliverable”
resources and ends in themselves. Our data set is just one resource from this process and more out-
comes are forthcoming.
One such multi-stage project is devoted to the free-list data that Purzycki, his PhD student
Theiss Bendixen, and Masters research assistants Cecilie Sandfeld and Andreas Damgaard are
working with to clean, code, and analyze. Participants listed things they thought angered and
pleased two different gods, things they thought constituted a good and bad person, and things
that please and anger the police, and various works (e.g., Bendixen & Purzycki, 2020; McNamara
& Purzycki, 2020; Purzycki & McNamara, 2016) have developed predictions to account for vari-
ation in those data. Lang and Purzycki also compiled and processed the preliminary ethnographic
interview data for public use and for integration with other databases (e.g., the Database for Reli-
gious History; Slingerland & Sullivan, 2017, and Seshat; Turchin et al., 2015) and are currently pre-
paring documentation for this sub-dataset.
Thinking about protocols, data sets, analytical code, and documentation as “deliverables”should
also entail a concomitant and unwavering commitment to data quality and project transparency. In
addition to data checks and audits, after finalizing our in-house analyses, we hired independent ana-
lysts (different across both waves) to reproduce our analyses to verify that they got the same results
using only our analytical scripts. We would recommend this practice for larger projects since com-
plex analytical scripts will involve typos and small changes that are easy to overlook can still dra-
matically change reported results. To combat persistent problems with scientific preservation and
transparency (see Hardwicke et al., 2020; Minocher et al., 2020), we embraced open scientific prac-
tices by providing all methodological protocols, data, and analytical code for use in both R and
STATA, and pre-registered the second wave analyses along with two other major projects. With
the publication of this special issue, we are making our cleaned data available, including new
data otherwise not used in our analyses. We hope researchers will make use of this resource,
10
and strongly encourage them to assess the data collection process, measures, and field sites contexts
carefully when they do.
Our examination of the same hypothesis across multiple sites is a reflection of the times; reveal-
ing and accounting for variation and diversity is far more taxing and less attractive to general audi-
ences than systematically testing in multiple places the same hypothesis about the role of powerful,
morally interested deities –like the Abrahamic god –play in human cooperation. As religion in its
myriad operationalizations was not our dependent variable, we had little to say empirically about
RELIGION, BRAIN & BEHAVIOR 205

accounting for it and how it might vary. The diversity of the samples was virtually unprecedented in
the general field of evolutionary and cognitive studies of religion, as we noted above. We hope that
such a sampling approach is adopted more widely in future research projects of this kind.
Nevertheless, we faced limitations in how to leverage the sampling diversity. We learned more
about the likelihood of a relationship between one class of deities and cooperation, but not about
the kinds of particular challenges and variations that exist in these sites and how that corresponds to
religious expression.
11
We relied on samples from 15 field sites to examine how widespread the
effect of moralistic gods on extended prosociality was, but we neither examined nor explained
the differences between populations that we found. The reported main effects were not present
in all samples and, in some cases, we actually found the reverse. Site-specific papers published in
the two special issues of Religion, Brain, & Behavior (see volume 8, issue 2) also provide insights
into the context of our particular sites but we did not fully investigate whether the site-specific
results may signal broader cultural and ecological patterns worth considering (e.g., birth rate, kin-
ship system, effective temperature, etc.). In a few papers (Baimel et al., present volume; Lang et al.,
2019; Purzycki et al., present volume; Vardy et al., present volume), we examined some distinctions
between Christian vs. non-Christian traditions, but given that our total site count was 15, some such
analyses are simply underpowered to treat this higher-order construct as a reliable predictor.
12
Future research of this scale should therefore carefully select field sites that do not substantially
vary on factors that may negate or amplify the effects under investigation, or select such a number
of sites that would make the exploration of these higher-order variables statistically feasible.
Furthermore, testing the same prediction with the same methods across a variety of societies is
not the only available approach to assess religiously inspired cross-community building. Another
approach would be to examine local instances of cooperation beyond kith and kin and examine
if and how religion is involved. Different forms of extended cooperative networks (e.g., see Bird
et al., 2019), along with the mechanisms of their maintenance are likely to reveal themselves
upon inspection. Indeed, casual scans of the ethnographic literature suggests this is the case (see
McNamara & Purzycki, 2020; Purzycki & McNamara, 2016). For example, one well-studied ritual
system–the Balinese water temple system–contributes to wide-scale cooperation in the form of
interconnected micro-societies (Lansing, 1987; Lansing & Kremer, 1993), but how beliefs play a
role has largely escaped empirical scrutiny. Another case includes the Hadza epeme secret society
that unites initiated men over many porous-but-disparate camps (Hill et al., 2014). Other cases
of religious ritual bridging multiple communities are manifold but anecdotal. Among the San
(Kalahari), for example, two antagonistic camps used a ritual context to soothe hostilities (Katz,
1982, p. 207). Clan-specific places of piety are open reciprocally to outside clan members to deposit
offerings to the spirits (Mongush, 1992, p. 169). The Ghost Dance of the American Great Plains and
Basin inspired a transnational resistance movement against European expansion (Mooney, 1896,
pp. 653–947), and the Sun Dance continues to bring many communities together.
Another critical aspect of religious systems that we did not address is how religious traditions
develop ontogenetically and the processes of cultural transmission (see Barrett, 2012; Harris
et al., 2006; Wen et al., 2020). While adults might have varying ideas about the degree to which
gods are moralistic, punitive, and knowledgeable, for instance, children might be taught in ways
that are (in)consistent with the portrait we offered here. For example, morality tales are cross-cul-
turally ubiquitous and involve gods in local pantheons. Moreover, recent work among hunter-gath-
erers has linked storytelling to band-level cooperation, as measured using behavioral experiments
(Smith et al., 2017). To the extent that they associate the supernatural with moral concern and the
consequences for breaching pan-cultural virtues and norms, such storytelling may have long-lasting
impacts on social development. This early introduction to the divine might bolster beliefs’adoption
later in life. Future research may therefore need to take into account also the individual history of
particular religious communities (Sosis, 2020).
Despite all the pitfalls we encountered along the way, we strongly believe that the future of the
study of religion, brain, and behavior should involve, among other fruitful approaches, large-scale,
206 B. G. PURZYCKI ET AL.

highly interdisciplinary teams collecting data of all kinds in diverse religious communities around
the globe. These teams can integrate cultural evolutionary theory with the best available methods
drawn from ethnography, history, psychology, sociology, economics, neuroscience, religious
studies, biology, and statistics. Here, we have offered a preliminary vision of one approach to
accomplish this vision. We hope that future researchers will be inspired by our insights, build on
our approaches, and learn from our mistakes. Good luck!
Notes
1. https://hecc.ubc.ca/hecc/cerc/.
2. https://philosophy.dept.shef.ac.uk/culture&mind/
3. http://jee.caltech.edu/research/experimental-economics/roots-human-sociality-phase-ii
4. For more details, see https://hecc.ubc.ca/the-cultural-evolution-of-prosocial-religions/.
5. These can be accessed here: https://osf.io/7y46w/.
6. In an exploratory interview (the “Religious Landscape Interview”’), participants were asked to freelist a maxi-
mum of five gods or spirits they could think of and then rank their importance in the participant’s life. From
these freelists, we drew out the most moralizing of the named deities and another important deity. We asked
the same questions about two different deities. Answers to these questions (e.g., How often do you think about
X deity?) were primarily on Likert scale options, but some were more open-ended and contingent on affirma-
tive responses to corollary questions (e.g., How does Y deity punish people?). All protocol materials can be
accessed here: https://github.com/bgpurzycki/Evolution-of-Religion-and-Morality/blob/master/CERC-Project-
2013_Protocols.zip.
7. In some cases, participant religion and therefore the identification of “co-religionists”was by default (e.g.,
everyone in a village is effectively Christian) whereas in other cases, participants were explicitly recruited
on the basis of a particular affiliation by direct questioning.
8. With this limitation in mind, it is worth pointing out that there was considerable variation across sites for the
role of local spirits on game outcomes. For example, some of our researchers documented stronger positive
effects for local spirit beliefs on game outcome than moralistic god beliefs (see Placek & Lightner, present
volume; Cohen et al., 2018). Some found that the more knowledge local spirits had, the more parochial indi-
viduals were toward other Buddhists (Purzycki & Kulundary, 2018). Others (McNamara & Purzycki, 2020)
show that by using various tablecloths as primes, participants playing on “ancestor spirit”cloths (with images
of traditional objects accompanied by the text “Fiji Islands Gift of Elders Soveniors [sic]”) versus “Jesus”(a
cloth quoting Mark 9:23, “Jesus said ‘All things are possible to him who believes’” with images of a crucifix
and Bible) has opposing effects on the breadth of cooperation. In this case, playing in the “Jesus”condition
expands the cooperative sphere to “church members from another island”while the “ancestor spirit”primes
fosters parochialism. Kundtová Klocová et al. (present volume) show that when individuals adopt a “sorcery
mode”rather than an “ancestor mode”of local spirit beliefs in Mauritius, participants are more likely to retain
money for themselves and local Hindu coreligionists in the random allocation games. The effect, however, did
not hold for dictator game outcomes.
9. A related issue stems from when questions do not correspond to the way participants think or talk (or are
willing to talk about). Often, they responded with “I don’t know,”which we subsequently recast as missing.
10. The data, all materials, and a catalogue of papers directly produced by the ERM project can be found here:
https://github.com/bgpurzycki/Evolution-of-Religion-and-Morality.
11. This lack of focus on diversity might have partly stemmed from the fact that in some cases, our field research-
ers were not specifically trained in the study of religion or had not conducted studies on the specific religious
traditions of their field sites. As noted, there was substantial variation in the prior experience of our team in
their field sites and studying religious beliefs and practices is one of the most challenging aspects of ethno-
graphic work. This is not to say that our site-specific papers have not contributed to more precise accounts
of some aspects of religion in the field sites. Indeed, the case of the Hadza—where Apicella (2018)offered
the first quantitative study of Hadza religion—requires us to rethink how we talk about “religion”in the
region. Note, too, that much of the Hadza protocol had to be altered in anticipation of their previous difficulty
with answering Likert scale items. Here, questions were reframed so that participants could choose between
“yes”,“no”, and “I don’t know”.
12. In a frequentist analytical framework at least, this poses a problem since variance would be calculated with too
few values (in this case two, as in Christian vs. non-Christian). In a Bayesian analytical framework, however,
variance is informed in part by a distribution set prior to introducing the model to the variance of the pooled
groups. In principle, then, something as basic as sex could (some suggest should) be treated as a higher-order
variable. The burden of effort, however, lies in carefully defining and examining one’s prior distributions of
intercepts and, if necessary, varying effects (see Gelman, 2006).
RELIGION, BRAIN & BEHAVIOR 207

Acknowledgments
The authors thank Adam Barnett for exhibiting abnormal amounts of patience and organization throughout the dur-
ation of this project, as well as the editors at Religion, Brain and Behavior for their continued encouragement and
management efforts. We thank Theiss Bendixen and our reviewers for their helpful feedback. A research grant,
“The Emergence of Prosocial Religions’from the John Templeton Foundation and a partnership grant (895-2011-
1009) from the Social Sciences and Humanities Research Council of Canada financially supported this project. Pur-
zycki acknowledges support from the Department of Human Behavior, Ecology, and Culture at the Max Planck Insti-
tute for Evolutionary Anthropology, the Aarhus University Research Foundation, and a Consequences of Formal
Education for Science and Religion Project grant funded by the Templeton Religion Trust and the Issachar Fund.
Joe Henrich thanks the Canadian Institute for Advanced Research.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Funding
This work was supported by Social Sciences and Humanities Research Council of Canada: [Grant Number 895-2011-
1009].
ORCID
Benjamin Grant Purzycki http://orcid.org/0000-0002-9595-7360
Martin Lang http://orcid.org/0000-0002-2231-1059
References
Achen, C. H. (2005). Let’s put garbage-can regressions and garbage-can probits where they belong. Conflict
Management and Peace Science,22(4), 327–339. https://doi.org/10.1080/07388940500339167
Apicella, C. L. (2018). High levels of rule-bending in a minimally religious and largely egalitarian forager population.
Religion, Brain & Behavior,8(2), 133–148. https://doi.org/10.1080/2153599X.2016.1267034
Argyle, M., & Beit-Hallahmi, B. (1975). The social psychology of religion. Routledge.
Atran, S., & Norenzayan, A. (2004). Religion’s evolutionary landscape: Counterintuition, commitment, compassion,
communion. Behavioral and Brain Sciences, 27(6), 713–770. https://doi.org/10.1017/S0140525X04000172
Azur, M., Stuart, E., Frangakis, C., & Leaf, P. (2011). Multiple imputation by chained equations: What is it and how does
it work? International Journal of Methods in Psychiatric Research,20(1), 40–49. https://doi.org/10.1002/mpr.329
Baimel, A. (2019). Reasoning about the supernatural: A cross-cultural examination of how and when intuitions shape
belief (Doctoral dissertation). https://open.library.ubc.ca/soa/cIRcle/collections/ubctheses/24/items/1.0379879.
Barrett, J. L. (2012). Born believers: The science of children’s religious belief. Simon and Schuster.
Barrett, H. C., Alexander, B., Crittenden Alyssa, N., Fessler Daniel, M. T., Simon, F., Michael, G., Joseph, H., Martin,
K., Geoff, K., Anne, P., Scelza Brooke, A., Stephen, S., Chris, V. R., Wanying, Z., & Stephen, L. (2016). Small-scale
societies exhibit fundamental variation in the role of intentions in moral judgment. Proceedings of the National
Academy of Sciences,113(17), 4688–4693. https://doi.org/10.1073/pnas.1522070113
Baumard, N., & Chevallier, C. (2015). The nature and dynamics of world religions: A life-history approach.
Proc. R. Soc. B,282(1818), 20151593. https://doi.org/10.1098/rspb.2015.1593
Bendixen, T., & Purzycki, B. G. (2020). Peering into the minds of gods: What cross-cultural variation in gods’con-
cerns can tell us about the evolution of religion. Journal for the Cognitive Science of Religion,5(2), 142–165. https://
doi.org/10.1558/jcsr.40951
Berger, P. L. (1967). The sacred canopy: Elements of a sociological theory of religion. Doubleday.
Bird, D. W., Bliege Bird, R., Codding, B. F., & Zeanah, D. W. (2019). Variability in the organization and size of hun-
ter-gatherer groups: Foragers do not live in small-scale societies. Journal of Human Evolution,131,96–108. https://
doi.org/10.1016/j.jhevol.2019.03.005
Boas, F. (1930). Some problems of methodology in the social sciences. In L. White (Ed.), The new social science (pp.
84–98). University of Chicago Press.
BorgerhoffMulder, M., Logsdon, W., & Barley, N. (1996). I’ve been gone far too long: Field study fiascoes and
expedition disasters. RDR Books.
Boyer, P. (2001). Religion explained: The evolutionary origins of religious thought. Basic Books.
208 B. G. PURZYCKI ET AL.

Boyer, P. (2020). Informal religious activity outside hegemonic religions: Wild traditions and their relevance to evol-
utionary models. Religion, Brain & Behavior,10(4), 459–472. https://doi.org/10.1080/2153599X.2019.1678518
Bulbulia, J., & Sosis, R. (2011). Signalling theory and the evolution of religious cooperation. Religion,41(3), 363–388.
https://doi.org/10.1080/0048721X.2011.604508
Bulbulia, J., Sosis, R., Harris, E., Genet, R., Genet, C., & Wyman, K. eds. (2008). The evolution of religion studies, the-
ories, & critiques. Collins Foundation Press.
Ensminger, J., & Henrich, J. (Eds.). (2014). Experimenting with social norms: Fairness and punishment in cross-cul-
tural perspective. Russell Sage Press.
Fischer, R., & Poortinga, Y. H. (2018). Addressing methodological challenges in culture-comparative research.
Journal of Cross-Cultural Psychology,49(5), 691–712. https://doi.org/10.1177/0022022117738086
Fowler, J. H., & Kam, C. D. (2007). Beyond the self: Social identity, altruism, and political participation. Journal of
Politics,69(3), 813–827. https://doi.org/10.1111/j.1468-2508.2007.00577.x
Geertz, A. W. (2020). How did ignorance become fact in American religious studies?: A reluctant reply to ivan
strenski. Studi E Materiali Di Storia Delle Religioni,86(1), 365–403.
Gelman, A. (2006). Prior distributions for variance parameters in hierarchical models (comment on article by
Browne and Draper). Bayesian analysis,1(3), 515–534. https://doi.org/10.1214/06-BA117A
Gelman, A., & Hill, J. (2007). Data analysis using regression and multilevel/hierarchical models. Cambridge University
Press.
Gray, K., Young, L., & Waytz, A. (2012). Mind perception is the essence of morality. Psychological Inquiry,23(2),
101–124. https://doi.org/10.1080/1047840X.2012.651387
Gurven, M., Zanolini, A., & Schniter, E. (2008). Culture sometimes matters: Intra-cultural variation in pro-social
behavior among Tsimane Amerindians. Journal of Economic Behavior & Organization,67(3-4), 587–607.
Hardwicke, T. E., Wallach, J. D., Kidwell, M. C., Bendixen, T., Crüwell, S., & Ioannidis, J. P. A. (2020). An empirical
assessment of transparency and reproducibility-related research practices in the social sciences (2014–2017). Royal
Society Open Science,7(2), 190806. https://doi.org/10.1098/rsos.190806
Harris, P., Pasquini, E. S., Duke, S., Asscher, J. J., & Pons, F. (2006). Germs and angels: The role of testimony in young
children’s ontology. Developmental Science,9(1), 76–96. https://doi.org/10.1111/j.1467-7687.2005.00465.x
Henrich, J. (2004). Cultural group selection, coevolutionary processes and large-scale cooperation. Journal of
Economic Behavior & Organization,53(1), 3–35. https://doi.org/10.1016/S0167-2681(03)00094-5
Henrich, J. (2009). The evolution of costly displays, cooperation and religion. Credibility enhancing displays and their
implications for cultural evolution. Evolution and Human Behavior,30(4), 244–260. https://doi.org/10.1016/j.
evolhumbehav.2009.03.005
Henrich, J., Bauer, M., Chytilova, J., Cassar, A., & Purzycki, B. G. (2019). War increases religiosity. Nature Human
Behavior,3(2), 129–135.
Henrich, J., Boyd, R., Bowles, S., Camerer, C., Fehr, E., Gintis, H., McElreath, R., Alvard, M. S., Barr, A., Ensminger, J.,
Henrich, N. S., Hill, K., Gil-White, F., Gurven, M., Marlowe, F. W., Patton, J. Q., & Tracer, D. (2005). “Economic
man”in cross-cultural perspective: Behavioral experiments in 15 small-scale societies. Behavioral and Brain
Sciences,28(6), 795–815. discussion 815-55. https://doi.org/10.1017/S0140525X05000142
Henrich, J., Ensminger, J., McElreath, R., Barr, A., Barrett, C., Bolyanatz, A., Cardenas, J. C., Gurven, M., Gwako, E.,
Henrich, N., Lesorogol, C., Marlowe, F., Tracer, D. P., & Ziker, J. (2010). Market, religion, community size and the
evolution of fairness and punishment. Science,327(5972), 1480–1484. https://doi.org/10.1126/science.1182238
Herrmann, B., Thöni, C., & Gächter, S. (2008). Antisocial punishment across societies. Science,319(5868), 1362–
1367. https://doi.org/10.1126/science.1153808
Hewlett, B. L. (2019). The secret lives of anthropologists: Lessons from the field.Routledge.
Hill, K. R., Wood, B. M., Baggio, J., Magdalena Hurtado, A., & Boyd, R. T. (2014). Hunter-gatherer inter-band inter-
action rates: Implications for cumulative culture. PLOS ONE,9(7), e102806. https://doi.org/10.1371/journal.pone.
0102806
Hruschka, D. J., Efferson, C., Jiang, T., Falletta-Cowden, A., Sigurdsson, S., McNamara, R., Sands, M., Munira, S.,
Slingerland, E., & Henrich, J. (2014). Impartial institutions, pathogen stress and the expanding social network.
Human Nature,25(4), 567–579. https://doi.org/10.1007/s12110-014-9217-0
Johnson, D. D. P. (2005). God’s punishment and public goods: A test of the supernatural punishment hypothesis in
186 world cultures. Human Nature,16(4), 410–446. https://doi.org/10.1007/s12110-005-1017-0
Kandler, A., & Powell, A. (2018). Generative inference for cultural evolution. Philosophical Transactions of the Royal
Society B: Biological Sciences,373(1743), 20170056. https://doi.org/10.1098/rstb.2017.0056
Katz, R. (1982). Boiling energy: Community healing among the Kalahari Kung. Harvard University Press.
Krátký, J., McGraw, J. J., Xygalatas, D., Mitkidis, P., & Reddish, P. (2016). It depends who is watching you: 3-D agent
cues increase fairness. PLoS ONE,11(2), 1–11. https://doi.org/10.1371/journal.pone.0148845
Kressel, N. J. (1990). Systemic barriers to progress in academic social psychology. The Journal of Social Psychology,
130(1), 5–27. https://doi.org/10.1080/00224545.1990.9922929
Kröll, M., & Rustagi, D. Reputation, honesty, and cheating in informal milk markets in India. (n.d.) SAFE Working
Paper No. 134, Available at SSRN: https://ssrn.com/abstract=2982365 or https://doi.org/10.2139/ssrn.2982365
RELIGION, BRAIN & BEHAVIOR 209

Lang, M., Purzycki, B. G., Apicella, C. L., Atkinson, Q. D., Bolyanatz, A., Cohen, E., Handley, C., Kundtová Klocová,
E., Lesorogol, C., Mathew, S., McNamara, R. A., Moya, C., Placek, C. D., Soler, M., Vardy, T., Weigel, J. L., Willard,
A. K., Xygalatas, D., Norenzayan, A., & Henrich, J. (2019). Moralizing gods, impartiality and religious parochial-
ism across 15 societies. Proceedings of the Royal Society B,286(1898), 20190202. https://doi.org/10.1098/rspb.2019.
0202
Lansing, J. S. (1987). Balinese ‘water temples’and the management of irrigation. American Anthropologist,89(2),
326–341. https://doi.org/10.1525/aa.1987.89.2.02a00030
Lansing, J. S., & Kremer, J. N. (1993). Emergent properties of Balinese water temple networks: Coadaptation on a
rugged fitness landscape. American Anthropologist, New Series,95(1), 97–114. https://doi.org/10.1525/aa.1993.
95.1.02a00050
McCauley, R. N. (2011). Why religion is natural and science is not. Oxford University Press.
McElreath, R. (2020). Statistical rethinking: A Bayesian course with examples in R and stan. 2nd edition. CRC Press.
McNamara, R. A., & Henrich, J. (2018). Jesus vs. the ancestors: How specific religious beliefs shape prosociality on
Yasawa Island, Fiji. Religion, Brain & Behavior,8(2), 185–204. https://doi.org/10.1080/2153599X.2016.1267030
McNamara, R. A., Norenzayan, A., & Henrich, J. (2016). Supernatural punishment, in-group biases, and material
insecurity: Experiments and ethnography from Yasawa, Fiji. Religion, Brain & Behavior,6(1), 34–55. https://
doi.org/10.1080/2153599X.2014.921235
McNamara, R. A., & Purzycki, B. G. (2020). Minds of gods and human cognitive constraints: Socio-ecological context
shapes belief. Religion, Brain and Behavior,10(3), 223–238. https://doi.org/10.1080/2153599X.2019.1678510
Minocher, R., Atmaca, S., Bavero, C., McElreath, R., & Beheim, B. (2020). Reproducibility of social learning research
declines exponentially over 63 years of publication. Preprint. PsyArXiv,1–20. https://doi.org/10.31234/osf.io/
4nzc7
Mongush, M. V. (1992). Lamaizm v tuve: Istoriko-etnograficheskoe issledovanie [lamaism in Tuva: Historical-ethno-
graphic research]. Kyzyl: Tuvinskoe Knizhnoe izdatel’stvo.
Mooney, J. (1896). The Ghost-dance religion. In J. W. Powell (director), Annual report of the bureau of ethnology to
the secretary of the smithsonian institution 14th Pt. 1 1892-1893 (pp. 653–947). Government Printing Office.
Nichols, A. D., Lang, M., Kavanagh, C., Kundt, R., Yamada, J., Ariely, D., & Mitkidis, P. (2020). Replicating and
extending the effects of auditory religious cues on dishonest behavior. PLoS ONE,15(8), e0237007. https://doi.
org/10.1371/journal.pone.0237007
Norenzayan, A. (2016). Theodiversity. Annual Review of Psychology,67(1), 465–488. https://doi.org/10.1146/
annurev-psych-122414-033426
Norris, P., & Inglehart, R. (2011). Sacred and secular: Religion and politics worldwide. (2nd edition) Cambridge
University Press.
Pasek, M. H., Shackleford, C., Smith, J. M., Vishkin, A., Lehner, A., & Ginges, J. (2020). God values the lives of my
out-group more than I do: Evidence from Fiji and Israel. Social Psychological and Personality Science,11(7), 1032–
1041.
Pearl, J., Glymour, M., & Jewell, N. P. (2016). Causal inference in statistics: A primer. John Wiley & Sons.
Pisor, A. C., Gervais, M. M., Purzycki, B. G., & Ross, C. T. (2020). Preferences and constraints: The value of economic
games for studying human behaviour. Royal Society Open Science,7(6), 192090. https://doi.org/10.1098/rsos.
192090
Purzycki, B. G. (2011). Tyvan cher eezi and the socioecological constraints of supernatural agents’minds. Religion,
Brain and Behavior,1(1), 31–45. https://doi.org/10.1080/2153599X.2010.550723
Purzycki, B. G. (2013). The minds of gods: A comparative study of supernatural agency. Cognition,129(1), 163–179.
Purzycki, B. G., Apicella, C., Atkinson, Q., Cohen, E., McNamara, R. A., Willard, A. K., Norenzayan, A., & Henrich, J.
(2016a). Cross-cultural dataset for the Evolution of Religion and Morality Project. Scientific Data,3(1), 16099.
https://doi.org/10.1038/sdata.2016.99
Purzycki, B. G., Apicella, C., Atkinson, Q. D., Cohen, E., McNamara, R. A., Willard, A. K., Xygalatas, D., Norenzayan,
A., & Henrich, J. (2016b). Moralistic gods, supernatural punishment and the expansion of human sociality.
Nature,530(7590), 327–330. https://doi.org/10.1038/nature16980
Purzycki, B. G., Henrich, J., Apicella, C., Atkinson, Q. D., Baimel, A., Cohen, E., McNamara, R. A., Willard, A. K.,
Xygalatas, D., & Norenzayan, A. (2018). The evolution of religion and morality: A synthesis of ethnographic and
experimental evidence from eight societies. Religion, Brain & Behavior,8(2), 101–132. https://doi.org/10.1080/
2153599X.2016.1267027
Purzycki, B. G., & Jamieson-Lane, A. (2017). AnthroTools: An R package for cross-cultural ethnographic data analy-
sis. Cross-Cultural Research,51(1), 51–74. https://doi.org/10.1177/1069397116680352
Purzycki, B. G., & Kulundary, V. (2018). Buddhism, identity, and class: Fairness and favoritism in the Tyva Republic.
Religion, Brain and Behavior,8(2), 205–226.
Purzycki, B. G., & Lang, M. (2019). Identity fusion, outgroups, and sacrifice: A cross-cultural test. Cognition,186,1–6.
https://doi.org/10.1016/j.cognition.2019.01.015
Purzycki, B. G., & McNamara, R. A. (2016). An ecological theory of gods’minds. In H. De Cruz, & R. Nichols (Eds.),
Advances in religion, cognitive science, and experimental philosophy (pp. 143–167). Continuum.
210 B. G. PURZYCKI ET AL.

Purzycki, B. G., Pisor, A. C., Apicella, C., Atkinson, Q., Cohen, E., Henrich, J., McElreath, R., McNamara, R. A.,
Norenzayan, A., Willard, A. K., & Xygalatas, D. (2018). The cognitive and cultural foundations of moral behavior.
Evolution and Human Behavior,39(5), 490–501. https://doi.org/10.1016/j.evolhumbehav.2018.04.004
Purzycki, B. G., Ross, C. T., Apicella, C., Atkinson, Q. D., Cohen, E., McNamara, R. A., Willard, A. K., Xygalatas, D.,
Norenzayan, A., Henrich, J., & van Elk, M. (2018). Material security, life history, and moralistic religions: A cross-
cultural examination. PLoS ONE,13(3), e0193856. https://doi.org/10.1371/journal.pone.0193856
Rustagi, D., Engel, S., & Kosfeld, M. (2010). Conditional cooperation and costly monitoring explain success in forest
commons management. Science,330(6006), 961–965. https://doi.org/10.1126/science.1193649
Saroglou, V., & Cohn, A. B. (2013). Cultural and cross-cultural psychology of religion. In Raymond F. Paloutzian & Crystal
L. Park (Eds.), Handbook of the Psychology of religion and spirituality, 2nd Ed (pp. 330–354). The Guilford Press.
Schulz, J. F., Bahrami-Rad, D., Beauchamp, J. P., & Henrich, J. (2019). The church, intensive kinship, and global
psychological variation. Science,366(6466), eaau5141. https://doi.org/10.1126/science.aau5141
Sears, D. O. (1986). College sophomores in the laboratory: Influences of a narrow data base on social psychology’s
view of human nature. Journal of Personality and Social Psychology,51(3), 515–530. https://doi.org/10.1037/0022-
3514.51.3.515
Shariff, A. F., Willard, A.K., Andersen, T., & Norenzayan, A. (2016). Religious priming: A meta-analysis with a focus on
prosociality. Personality and Social Psychology Review,20(1), 27–48. https://doi.org/10.1177/1088868314568811
Silk, M. J., Harrison, X. A., & Hodgson, D. J. (2020). Perils and pitfalls of mixed-effects regression models in biology.
PeerJ,8(August), e9522. https://doi.org/10.7717/peerj.9522
Slingerland, E., & Collard, M. (Eds.). (2011). Creating consilience: Integrating the sciences and the humanities. Oxford
University Press.
Slingerland, E., & Sullivan, B. (2017). Durkheim with data: The database of religious history. Journal of the American
Academy of Religion,85(2), 312–347. https://doi.org/10.1093/jaarel/lfw012
Smith, D., Schlaepfer, P., Major, K., Dyble, M., Page, A. E., Thompson, J., Chaudhary, N., Salali, G. D., Mace, R.,
Astete, L., Ngales, M., Vinicius, L., & Migliano, A. B. (2017). Cooperation and the evolution of hunter-gatherer
storytelling. Nature Communications,8(1) 1–9. https://doi.org/10.1038/s41467-017-02036-8
Sosis, R. (2020). The last Talmudic demon? The role of ritual in cultural transmission. Philosophical Transactions of the
Royal Society of London. Series B, Biological Sciences,375(1805), 20190425. https://doi.org/10.1098/rstb.2019.0425
Sosis, R., & Alcorta, C. (2003). Signaling, solidarity, and the sacred: The evolution of religious behavior. Evolutionary
Anthropology: Issues, News, and Reviews,12(6), 264–274. https://doi.org/10.1002/evan.10120
Sosis, R., & Bressler, E. R. (2003). Cooperation and commune longevity: A test of the costly signaling theory of reli-
gion. Cross-Cultural Research,37(2), 211–239. https://doi.org/10.1177/1069397103037002003
Stark, R. (2002). Physiology and faith: Addressing the “universal”gender difference in religious commitment. Journal
for the Scientific Study of Religion,41(3), 495–507. https://doi.org/10.1111/1468-5906.00133
Turchin, P., Brennan, R., Currie, T. E., Feeney, K. C., Francois, P., Hoyer, D., & Manning, J. G. (2015). Seshat: The
global history databank. http://seshatdatabank.info/
van Buuren, S., & Groothuis-Oudshoorn, K. (2011). mice: Multivariate imputation by chained equations. Journal Of
Statistical Software,45(3), 1–67. https://doi.org/10.1177/0962280206074463
van Elk, M., Matzke, D., Gronau, Q. F., Guan, M., Vandekerckhove, J., & Wagenmakers, E.-J. (2015). Meta-analyses
are no substitute for registered replications: A skeptical perspective on religious priming. Frontiers in Psychology,
6, 1365. https://doi.org/10.3389/fpsyg.2015.01365
Vardy, T. (2019). A test of functional theories of religion in a non-western sample (Doctoral dissertation). https://
researchspace.auckland.ac.nz/bitstream/handle/2292/47731/whole.pdf?sequence=2.
Wen, N. J., Willard, A. K., Caughy, M., & Legare, C. H. (2020). Watch me, watch you: Ritual participation increases
in-group displays and out-group monitoring in children. Philosophical Transactions of the Royal Society B,375
(1805), 20190437. https://doi.org/10.1098/rstb.2019.0437
White, C. J., Kelly, J. M., Shariff, A. F., & Norenzayan, A. (2019). Supernatural norm enforcement: Thinking about
karma and God reduces selfishness among believers. Journal of Experimental Social Psychology,84, 103797. https://
doi.org/10.1016/j.jesp.2019.03.008
Wiessner, P. (2009). Experimental games and games of life among the Ju/’hoan bushmen. Current Anthropology,50
(1), 133–138. https://doi.org/10.1086/595622
Wildman, W. J., & Sosis, R. (2011). Stability of groups with costly beliefs and practices. Journal of Artificial Societies
and Social Simulation,14(3), 1–14. https://doi.org/10.18564/jasss.1781
Wilson, D. S. (2010). Darwin’s cathedral: Evolution, religion, and the nature of society. University of Chicago Press.
Wilson, R. C., & Collins, A. G. E. (2019). Ten simple rules for the computational modeling of behavioral data. Edited
by Timothy E Behrens. ELife,8(November), e49547. https://doi.org/10.7554/eLife.49547
Woodbury, R. (2012). The missionary roots of liberal democracy. American Political Science Review,106(2), 244–274.
https://doi.org/10.1017/S0003055412000093
Yenni, G. M., Christensen, E. M., Bledsoe, E. K., Supp, S. R., Diaz, R. M., White, E. P., & Morgan Ernest, S. K. (2019).
Developing a modern data workflow for regularly updated data. PLOS Biology,17(1), e3000125. https://doi.org/10.
1371/journal.pbio.3000125
RELIGION, BRAIN & BEHAVIOR 211