Christianity spread faster in small, politically structured societies

Abstract

Over the past 2,000 years, Christianity has grown from a tiny Judaic sect to the world’s largest religious family¹. Historians and social scientists have long debated whether Christianity spread through a top-down process driven by political leaders or a bottom-up process that empowered social underclasses2–6. The Christianization of Austronesian populations is well-documented across societies with a diverse range of social and demographic conditions⁷. Here, we use this context to test whether political hierarchy, social inequality and population size predict the length of conversion time across 70 Austronesian cultures. We also account for the historical isolation of cultures and the year of missionary arrival, and use a phylogenetic generalized least squares method to estimate the effects of common ancestry and geographic proximity of cultures⁸. We find that conversion to Christianity typically took less than 30 years, and societies with political leadership and smaller populations were fastest to convert. In contrast, social inequality did not reliably affect conversion times, indicating that Christianity’s success in the Pacific is not due to its egalitarian doctrine empowering social underclasses. The importance of population size and structure in our study suggests that the rapid spread of Christianity can be explained by general dynamics of cultural transmission.

Full text

Letters
https://doi.org/10.1038/s41562-018-0379-3
1Max Planck Institute for the Science of Human History, Jena, Germany. 2Social and Evolutionary Neuroscience Research Group, Department of
Experimental Psychology, University of Oxford, Oxford, UK. 3School of Psychology, University of Auckland, Auckland, New Zealand. 4School of Humanities,
Faculty of Arts, University of Auckland, Auckland, New Zealand. 5Research School of the Social Sciences, Australian National University, Canberra,
Australian Capital Territory, Australia. 6Institute of Cognitive and Evolutionary Anthropology, University of Oxford, Oxford, UK. *e-mail: watts@shh.mpg.de
Over the past 2,000 years, Christianity has grown from a tiny
Judaic sect to the world’s largest religious family1. Historians
and social scientists have long debated whether Christianity
spread through a top-down process driven by political lead-
ers or a bottom-up process that empowered social under-
classes2–6. The Christianization of Austronesian populations
is well-documented across societies with a diverse range of
social and demographic conditions7. Here, we use this context
to test whether political hierarchy, social inequality and pop-
ulation size predict the length of conversion time across 70
Austronesian cultures. We also account for the historical iso-
lation of cultures and the year of missionary arrival, and use
a phylogenetic generalized least squares method to estimate
the effects of common ancestry and geographic proximity of
cultures8. We find that conversion to Christianity typically
took less than 30 years, and societies with political leadership
and smaller populations were fastest to convert. In contrast,
social inequality did not reliably affect conversion times, indi-
cating that Christianity’s success in the Pacific is not due to
its egalitarian doctrine empowering social underclasses. The
importance of population size and structure in our study sug-
gests that the rapid spread of Christianity can be explained by
general dynamics of cultural transmission.
According to top-down theories of conversion, political leaders
such as Roman Emperor Constantine the Great played a major role
in Christianity’s success3,9. Consistent with this theory, Constantine
is known to have placed Christians in positions of power, encour-
aged conversion among the population and supported the church
through financial aid6. In contrast, bottom-up theories of conver-
sion argue that Christianity spread because its relatively egali-
tarian doctrines appealed to majority underclasses, rather than
through an elite minority2,10,11. Advocates of this theory suggest
that Constantine converted once Christianity had already become
a major social force and that his support for Christianity was a stra-
tegic move to appease the masses6,10. This debate underscores the
potential importance of both political authority and social inequal-
ity in the history of Christianity. However, these effects are diffi-
cult to disentangle from the historical record because both social
inequality and centralized political control tended to occur together
in early Christian expansions.
Despite the importance of demography in general processes of
cultural evolution, few historical or social scientific models of reli-
gious change discuss population size. Experimental and simulation-
based studies of cultural evolution show that larger populations with
better connected sub-populations can generate cultural innovations
at higher rates, transmit innovations between sub-populations
and are better able to retain cultural innovations than smaller pop-
ulations with less connected sub-populations12–17. However, many
of these models also show that it can take longer for new traits to
spread through larger populations than smaller populations17.
Together, the literature suggests that the rate of religious change
could be slower in larger and less connected populations than in
smaller and more connected populations18. Given that population
size can affect cultural transmission and co-varies with social struc-
ture, any attempt to systematically evaluate the drivers of the spread
of Christianity needs to account for the effects of social structure
and population size simultaneously.
Although Christianity is now found across all world regions1,
Austronesian-speaking peoples were among the most recent to
convert, and their histories of conversion are relatively well-doc-
umented. Austronesian contact with Christianity typically first
occurred through the arrival of European explorers and mission-
aries between the sixteenth and nineteenth centuries. This con-
tact was documented in written records and incudes details of the
population sizes and social structures of Austronesian cultures, as
well as their processes of conversion to Christianity7,19. At the time
of European contact, Austronesian societies ranged in size from
fewer than one hundred to hundreds of thousands, and their social
systems ranged from egalitarian, family-based communities such
as the Isneg, up to hierarchically organized political states such as
Hawaii7,20. Due to the richly documented conversion process and
their diverse social systems, Austronesian cultures provide a natu-
ral experiment for testing how population size and social structure
affect the cultural transmission of Christianity19.
Here, we test three hypotheses about the effects of population
size and social structure on the transmission of Christianity in
Austronesian cultures. First, we test whether cultures with greater
political organization are faster to convert to Christianity, as pre-
dicted by top-down theories of conversion. Political complexity
refers to the number of levels of political hierarchy in a culture and
provides a measure of the influence of leaders over a population21.
Political organizations include chiefdoms, kingdoms and state gov-
ernments, which all have the potential to centralize authority and
unite otherwise separate local communities. Because political com-
plexity also provides a measure of how connected the local com-
munities of a culture are, it can be predicted to facilitate conversion,
as shown by existing literature on population structure in cultural
evolution12,14. Second, we test whether cultures with higher levels
of social inequality are faster to convert, as predicted by bottom-
up theories of conversion. Social inequality—also known as social
stratification—refers to inherited differences in wealth and sta-
tus within a culture22. Examples of social inequality include the
Christianity spread faster in small, politically
structured societies
Joseph Watts1,2*, Oliver Sheehan1,3, Joseph Bulbulia1,4, Russell D. Gray1,3,5 and Quentin D. Atkinson 1,3,6
NATURE HUMAN BEHAVIOUR | www.nature.com/nathumbehav

Letters Nature HumaN BeHaviour
existence of noble classes that inherit land and have priority access
to resources, as well as slave classes who are strictly born into subju-
gation. While social inequality involves the structuring of a popula-
tion into separate classes, it does not necessarily connect different
local communities within a culture. Third, we test whether larger
populations are slower to convert, as predicted by general processes
of cultural diffusion throughout populations17,18. We also include
two additional control variables: one on cultural isolation and the
other on the year of missionary arrival. Cultural isolation refers to
the geographic distance from the nearest neighbouring society and
controls for historical exposure to different social groups. The year
of missionary arrival refers to the first year of concerted missionary
effort and is included to control for changes in the rates of conver-
sion over the three centuries that missionaries arrived.
We use a phylogenetic generalized least squares spatial (PGLS-
spatial) method8 to test how political complexity, social inequality,
population size, cultural isolation and year of missionary arrival pre-
dict the length of conversion time in 70 Austronesian cultures. By
incorporating matrices representing how recently cultures shared
a common ancestor, as well as their geographic distance from one
another, PGLS-spatial enables us to; (1) test whether close cultural
relatives within the Austronesian family tended to have similar con-
version times, suggesting the existence of other culturally inherited
traits that influence conversion time; (2) test whether geographic
neighbours tended to have similar conversion times, suggesting that
conversion time is influenced by other unmodelled cultural or eco-
logical factors that vary geographically; and (3) statistically adjust
for any such dependencies between cultures23,24.
The first year of concerted missionary efforts in our sample
ranged from 1668 to 1950, with a median of 1865.00 and a mean
of 1868.83 (s.d. = 48.15). The year in which at least half the popula-
tion had converted to Christianity ranged from 1695 to 1987, with a
median of 1902.50 and a mean of 1899.29 (s.d. = 51.54). The length
of this conversion time ranged from less than 1 year to 203 years with
a median of 25.00 years and a mean of 30.46 years (s.d. = 31.79). Of
the 70 cultures sampled, 11 had no centralized political authority, 13
had political authority within the local community, 33 had 1 level
of political authority beyond the local community, 9 had 2 levels of
political authority beyond the local community and 4 had 3 levels
of political authority beyond the local community (Supplementary
Table 1). For our variable on social inequality, we found that 13 cul-
tures were egalitarian, 37 had moderate social stratification and 20
had high levels of social stratification. Population sizes at the mid-
point of conversion ranged from 62 to 500,000 with a median of
5,800 and mean of 38,443.46 (s.d. = 79,843.81).
To provide a complete model of the effects of population size
and structure on conversion times, we ran full multivariate PGLS-
spatial analyses that included all predictor and control variables.
In this full model, population size was found to be significantly
positively correlated with conversion times, indicating that larger
populations took longer to convert to Christianity. Consistent with
the top-down theory, political complexity was negatively associated
with conversion times (Table 1). Counter to the bottom-up theories,
there was no reliable support for an association between conversion
time and social inequality (Table 1). Cultural isolation was found
to significantly negatively predict conversion times, indicating that
more isolated cultures tended to be faster to convert to Christianity
than less isolated cultures. The year of missionary arrival also sig-
nificantly negatively predicted conversion times, indicating that
societies in which missionaries arrived more recently were faster to
convert to Christianity. The lack of support for phylogenetic and
spatial dependencies is consistent with the lack of geographic and
phylogenetic clustering evident in the length-of-conversion-time
variable (Fig. 1 and Supplementary Table 2). This suggests that
shared cultural ancestry and geographically clustered factors had
little effect on the length of conversion time in Austronesia.
We note that the two cultures that took the longest to convert in
this study—Ifuagao and Iban—lacked any form of political orga-
nization (Supplementary Fig. 1). To assess how robust the effects
of political complexity were, we re-ran the multivariate analyses
removing these cultures and the 9 other cultures that lacked any
form of political structure (political complexity = 0). In the result-
ing model, political complexity was no longer a significant predic-
tor of conversion times (Supplementary Table 3). This indicates
that the significant effects of political complexity on conversion
times in our full model may be due to societies that lack politi-
cal organization taking longer to convert, rather the number of
levels of political hierarchy having a general effect on conversion
times. Population size remained a significant predictor of conver-
sion times even after societies without political organization were
excluded from our analyses.
Our findings show the importance of population size in cul-
tural transmission, and suggest that political organization, but not
social inequality, also played a role in the spread of Christianity. In
Austronesian-speaking cultures, the conversion process occurred
remarkably rapidly, with the majority of cultures sampled taking
less than 30 years to convert. Conversion appears to have occurred
mainly through the efforts of foreign missionaries, and cultures’
reactions to missionaries ranged from highly hospitable to violently
hostile. For example, on Kapingamarangi, missionaries arrived after
a period of famine and the entire culture converted p eacefully within
1 year25. In contrast, it took missionaries around 65 years to convert
half the Kwaio population; multiple missionaries were killed in the
process, and today around one-third of the population retain their
earlier beliefs and remain unaffiliated with Christianity26,27.
One reason that cultures with political organization were faster
to convert to Christianity may be that leaders imposed Christianity
on populations through top-down control, as previous theorists
have speculated3,9. A common strategy for missionaries was to
focus on winning over political leaders by providing them with
material goods and access to trade opportunities. For example, the
standard approach of the prolific and highly successful mission-
ary John Williams was to “put the initial resources of the mission
at the disposal of the most powerful chief he could win over”28. In
many Austronesian cultures, political leaders held great authority,
Table 1 | Results of multivariate PGLS-spatial analysis with all
variables included
Predictor βs.e. Pvalue
Political complexity –9.93** 3.71 0.009
Social inequality 1.48 5.99 0.805
Log10[population size] 15.15** 4.74 0.002
Log10[cultural isolation] –7.53* 3.75 0.049
Year of missionary arrival –0.21** 0.08 0.009
Intercept 397.17* 159.89 0.015
Dependency Mean effect P value
Phylogeny (λ′ ) 0 1.000
Spatial (φ) 0 1.000
Independent (y) 1 –
Analyses were replicated across a sample of 4,200 trees and the results of each replication were
identical. The upper section reports the estimated β values of each of the predictor variables,
along with their s.e. and P values. The lower section represents the relative effects of phylogeny
(λ′ ) and geographic distance (φ) in the model with the maximum likelihood, as well as the effects
independent of phylogeny and geographic distance (y). The P value for the effect of λ′ indicates
whether the best-fitting full model is significantly different from the best-fitting model in which
λ′ is constrained to 0. The P value for the effect of φ indicates whether the best-fitting full model
is significantly different from the best-fitting model in which φ is constrained to 0. *P < 0.05;
**P < 0.01.
NATURE HUMAN BEHAVIOUR | www.nature.com/nathumbehav

Letters
Nature HumaN BeHaviour
and missionaries might have aimed to exploit their influence and
power. However, it is also important to recognize that the life of a
missionary was a vulnerable and risky one and even Williams was
eventually killed and eaten on the island of Erromango. An alterna-
tive explanation for missionaries’ focus on winning over political
leaders is that this was simply a strategy of self-preservation.
Another possible reason that politically organized cultures were
faster to convert is that political structures provided connections
between what would otherwise have been separate, potentially con-
flicting communities within a culture. Instead of missionaries hav-
ing to seed Christianity anew in each community, Christianity could
be transmitted directly between the different communities of politi-
cally structured societies. This explanation fits with previous studies
showing the importance of connectivity between sub-populations
in processes of cultural evolution12,14,15. However, it is worth noting
that we found only a modest effect size of political complexity on
conversion time, and this effect was probably driven by a relatively
small number of cultures that lacked political organization within
or between communities. This suggests that what is important in
our study is whether a culture has political leaders at all, rather than
the number of levels of political hierarchy it possesses.
We do not find evidence that the presence of an underclass in
more stratified societies facilitated the spread of Christianity via
a bottom-up process of conversion. Historical records show that
even when missionaries did not focus their efforts on political lead-
ers, there was little to suggest that Christianity spread because it
empowered social underclasses or appealed to ‘slave morality’11. For
example, in the case of Nias, rapid conversions of the broad popula-
tion occurred after around 65 years of extensive work by mission-
aries29. The eventual conversion was attributed to the missionary
Ruderdorf’s threat of “expulsion from the Last Supper” for those
who did not repent29. This threat resonated with the ultimate and
socially damaging Nias snub of excluding someone from the divi-
sion of meat at feasts29.
While not central to the hypotheses of this study, our control
variables on cultural isolation and year of missionary arrival were
both found to be significant predictors of conversion times. One
possible reason that more isolated cultures were faster to convert
is that contact with other cultures can involve conflict, potentially
resulting in societies with greater contact being less receptive to
foreigners. One possible reason that cultures with more recent
missionary arrivals were faster to convert is that over time mis-
sionaries improved their conversion techniques and knowledge of
Austronesian languages, customs and social structures. A second
potential reason that cultures became faster to convert is that mis-
sionaries often provided access to trade opportunities and educa-
tion. Once cultures had witnessed other societies gaining these
benefits, they may have been more willing to convert. A third poten-
tial reason that cultures became faster to convert is that Christianity
was also sometimes directly transmitted between Austronesian
societies. For example, after being dropped off with only the clothes
on his back and some untranslated literature, the Tahitian mission-
ary Papeiha “accomplished more in Rarotonga in two years than the
English missionaries in Tahiti had in twenty”30. Once Christianity
was introduced, Austronesian peoples also played an important role
in the spread of Christianity in the Pacific.
Population size was the strongest predictor of conversion times
in our study, with larger populations taking significantly longer
to convert than smaller populations. The effect of population size
alone explained almost one-quarter of the variation in conversion
times and held across all of our analyses. While individual-level data
on the spread of Christianity within Austronesian populations are
limited, studies of Western populations show that one of the stron-
gest predictors of whether an individual converts is the number of
ties they have to existing converts31. This suggests that Christianity
can spread by cultural transmission between general community
members, and that individuals decide whether to convert based on
whether those around them have converted. This process of conver-
sion can result in frequency-dependent transmission32, whereby the
probability that any given individual converts is proportional to the
frequency of Christianity within the population. This can be con-
trasted with processes of conversion that do not involve frequency-
dependent transmission, such as if missionaries simply arrived and
broadcasted their message to the entire population, and individu-
als decided to convert based on their own personal characteristics.
Likewise, frequency-dependent transmission is not necessarily
100° E 120° E 140° E 160° E 180° 160° W 140° W 120° W
40° S
20° S
0°
20° N
200
150
100
50
0
Conversion time (years)
Latitude
Longitude
Fig. 1 | Number of years taken to convert for each of the 70 Austronesian cultures sampled. The box in the lower left corner represents Madagascar.
Image created using a figure from Natural Earth.
NATURE HUMAN BEHAVIOUR | www.nature.com/nathumbehav

Letters Nature HumaN BeHaviour
involved if political leaders convert first and then simply impose
Christianity on the general population. While Christianity is likely
to have spread through a variety of processes, the population size
effect we identify may suggest that frequency-dependent processes
of cultural transmission played an important role.
One of the features of frequency-dependent cultural transmis-
sion is that it tends to take longer for traits to spread through larger
populations than smaller populations17,33. When a novel trait is
seeded to a larger population, it starts off at a lower relative fre-
quency, and so a smaller proportion of the population come into
contact with that trait and subsequently adopt it. Under such a
scenario, when Christianity is initially introduced to a popula-
tion, individuals have on average fewer ties to other Christians
in larger populations than in smaller populations. As the propor-
tion of the population that are converted increases, so too does
the average number of ties that individuals have to Christianity,
and subsequently how likely they are to convert. To the extent
that frequency-dependent cultural transmission is important for
the conversion process, Christianity is expected to take longer to
gain a majority in larger populations than smaller populations,
but can rapidly spread once its frequency has increased within the
population. This relatively simple process of cultural transmis-
sion may help to explain how Christians in Ancient Rome num-
bered in the thousands in the year 100, and the tens of millions in
the year 3504,6.
While frequency-dependent cultural transmission can help
explain the importance of population size in the conversion pro-
cess, and also how the spread of Christianity can gain momen-
tum even in larger populations, this relies on Christianity having
the propensity to be transmitted between individuals in the first
place. Differences between the features of Christianity and the
features of ethnic religious systems are likely to provide part of the
explanation for Christianity’s tendency to spread. Some pertinent
features of Christianity include a universal rather than ethnocen-
tric doctrine, a central moralistic high God rather than a pan-
theon of deities, encouragement of proselytization and fertility,
and the claim of exclusive truth34–36. Another part of Christianity’s
remarkable success in the Pacific is its historical link to colonial
powers34–36. European colonists did not just bring Christianity
with them; they also brought diseases that decimated populations,
novel resources and sophisticated weaponry that may all have lent
credibility to Christianity37.
Population size is regarded as an important factor in the evo-
lution of cultural complexity13,15,16,38, but existing cross-cultural
research shows an inconsistent relationship between population size
and measures of cultural complexity39–44. One potential explanation
for this inconsistency is that the effects of population size vary across
different processes of cultural evolution38. From a macroevolution-
ary perspective, there are three major pathways through which cul-
tures acquire cultural innovations: independent innovation within a
culture, inheritance from a common ancestor and the borrowing of a
trait from a different cultural group45,46. Existing research shows that
greater population size can increase the rate at which cultural inno-
vations are generated and retained within cultures13,15. However, our
study shows that greater population sizes can also slow the spread
of traits throughout cultures, and may even inhibit the adoption of
foreign innovations. This suggests that the effects of population size
on the evolution of cultural complexity could depend on the avail-
able sources of cultural innovations14,38. When cultures are relatively
isolated and innovations must be generated within cultures, greater
population size will facilitate the evolution of cultural complexity.
When contact between cultures is common and innovations are
predominantly acquired through borrowing, greater population is
less important as an engine of innovation and may even inhibit the
adoption of new cultural variants. Over human history, the relative
importance of these different sources of cultural innovation is likely
to have changed, suggesting a complex role of population size in the
evolution of cultural complexity.
The lack of any phylogenetic or geographic signal in the conver-
sion time data implies that conversion time was not reliably affected
by recent cultural ancestry, diffusion between cultural groups
or regional ecology. This finding also suggests that the dynamics
observed in our data might be quite general. Outside Austronesia,
there are similar patterns of conversion in the histories of Africa
and Latin America1 as missionaries worked throughout these
regions to rapidly replace ethnic religions with Christianity. The
spread of Christianity in these regions has been its largest source
of growth over recent centuries, and explains how it has become
the dominant world religion today1. Here, we have shown that the
rapid spread of Christianity can be explained by general dynamics
of cultural transmission.
Methods
We sampled all cultures from the Pulotu database for which there was sufficient
ethnographic and historical information on the variables of interest. Pulotu is a
database that documents the religious and social features of 116 Austronesian
cultures7. The term ‘Austronesian’ refers to a language-based grouping of cultures
that originated in Taiwan47, spread west to Madagascar, east to Rapa Nui (Easter
Island) and south to Aotearoa (New Zealand)—an area covering over half the
world’s longitude and one-third of its latitude. The majority of Austronesian
societies relied primarily on horticulture and/or agriculture for subsistence,
although there was variation in the intensity of agriculture as well as the
importance of trade, fishing, hunting and gathering across societies7,48. Despite
a history of contact with Muslim and Hindu traders in parts of Southeast Asia,
many Austronesian cultures remained isolated from major world religions until
the arrival of Europeans in the past few hundred years. Austronesian cultures often
retain elements of their traditional religious systems, but today all (or almost all)
cultures in the Pulotu database have largely converted to a major world religion.
As we were interested in testing theories about the spread of Christianity, cultures
that converted to Islam, Hinduism or Buddhism were excluded from the study.
This resulted in a sample of 70 Austronesian cultures, covering all major regions
of Austronesia. The variables on political complexity, social inequality, cultural
isolation and geographic location are based on the traditional state of the cultures,
before colonization and conversion to Christianity. Data were gathered from the
records of missionaries, explorers and ethnographers, as well as official government
censuses (Supplementary Table 1). Each culture was coded by two trained coders.
The first coder collected data and coded the variables, while the second coder
reviewed the first coder’s coding decisions and searched for additional resources. In
cases of initial difference, the coders reviewed the materials together and came to
full agreement.
We coded political complexity based on Murdock and Provost’s
‘level of political integration’ coding scheme21. Cultures were coded
as 0 if there was no centralized political authority within the culture, 1
if there was a centralized political authority within the local community but
no political authority beyond the local community, 2 if there was one level of
political authority beyond the local community, 3 if there were two levels of
political authority beyond the local community, and 4 if there were three or
more levels of political authority beyond the local community. When the level of
political complexity varied within a culture, the culture was assigned the highest
observed level.
We coded social inequality using a variable inspired by Sahlins’ classification
scheme of social stratification22. Cultures were coded as 1 if they were egalitarian
(meaning there was no or minimal potential for wealth and/or status to be
inherited between generations), 2 if they were moderately stratified (meaning
there were inherited intergenerational differences in wealth and/or status but
these differences were minor, not clearly delineated or could change within one’s
lifetime) and 3 if they were highly stratified (meaning there were intergenerational
differences in wealth and/or status that were associated with pronounced
differences in living standards, clearly delineated and could not be changed within
one’s lifetime).
We defined the length of the conversion process as the number of years
between the onset of missionization and the year in which at least half the
population had converted to Christianity. The onset of missionization was defined
as the first year in which missionaries arrived and made a concerted effort to gain
converts. Brief missionary trips aimed at researching the size of communities and
their reaction to foreigners were not counted as attempts at missionization. We
defined the year of conversion as the year in which at least half of the population
identified with a Christian denomination. When the year at which half of
the population had converted was not documented, we estimated the year of
conversion based on the nearest known data points and a linear increase in the
proportion of the population converted over time. For example, if demographic
records indicated that 45% of the population had converted in the year 1850
NATURE HUMAN BEHAVIOUR | www.nature.com/nathumbehav

Letters
Nature HumaN BeHaviour
and 60% of the population had converted in 1859, we would estimate the year of
conversion to be 1853.
As a measure of population size, we used demographic records to estimate
the number of people in the culture at the mid-point between the first year
of concerted missionary effort and the year at which half the population had
converted to Christianity. When data on the population size were not available
for the year of interest, we estimated the population size based on the nearest
known data points and an exponential model of population growth. For
example, if a population numbered 12,000 in the year 1830 and 15,000 in the
year 1860, we would estimate that the population size was 14,000 in the year
1850 (rounded to avoid spurious accuracy). As population sizes in our sample
were highly skewed to the lower end, we used the log10 of the population size in
all analyses.
For our variable on cultural isolation, we coded the distance to the nearest
landmass inhabited by a different cultural group in kilometres. When there were
other cultures on the same landmass as the culture being coded, the nearest
neighbour was coded as 0. As the distances were highly skewed to the lower end,
we used the log10 of the distance in all analyses. This variable on cultural isolation
provides a proxy for the relative historical isolation of Austronesian societies.
To test for phylogenetic signal, we calculated Blomberg’s K for each variable
used in this study49–51. Blomberg’s K was calculated across a sample of 4,200 trees
and we report the mean K and P values here (see Supplementary Table 2 for further
details). We found no significant phylogenetic patterning for log10 population size
(K = 0.60, P = 0.097) or length of conversion time (K = 0.52, P = 0.256). However,
we found that political complexity (K = 0.54, P = 0.017), social inequality (K = 0.63,
P = 0.013), cultural isolation (K = 2.61, P < 0.001) and the year of missionary arrival
(K = 1.06, P = 0.003) showed evidence of phylogenetic patterning, indicating that
there are potential dependencies in our data related to the common ancestry of the
cultures sampled.
Standard regression methods assume that cultures are independent from
one another, despite them being related through common descent and patterns
of borrowing23,52 (Supplementary Table 2). This non-independence can lead
to spurious correlations, and the difficulty of distinguishing such spurious
correlations from correlations that result from actual causal relationships between
variables has come to be known as Galton’s problem35. The PGLS-spatial method
developed by Freckleton and Jetz8 makes it possible to address Galton’s problem
using a phylogeny to control for non-independence due to common ancestry, and
geographic proximity to control for non-independence due to diffusion between
cultures. Model-based approaches to quantifying non-independence have recently
been used to identify the effects of cultural ancestry and geography on political
unrest across Indo-European-speaking nations24, deforestation in the Pacific53
and human development indicators across Eurasian countries54. Here, we use
this PGLS-spatial method to simultaneously estimate the effects of our predictor
variables, cultural ancestry and spatial diffusion on the length of conversion. Two
series of additional PGLS-spatial analyses are also reported in the Supplementary
Materials. The first tests the bivariate relationship between each predictor variable
and the conversion time variable, while the second tests how potential collinearity
of social inequality and political complexity effects our results (Supplementary
Notes and Supplementary Tables 4–10).
The PGLS-spatial method includes two covariance matrices in the model
structure—one providing the expected covariance structure due to phylogenetic
history and the other providing the expected covariance relationships due to
geographic proximity8. To create the covariance matrix of phylogenetic distance,
we matched each culture in our sample to an established language-based family
tree built using basic vocabulary data47 (see Supplementary Table 1 for the
correspondence between cultural and linguistic groups). Basic vocabulary items,
such as kinship terms, pronouns and numbers, have low rates of borrowing
between languages, making them a good marker of cultural history55. For the
phylogenetic distance matrix, the distance between cultures was calculated
based on the distance from their most recent common ancestor in the language
phylogeny. To account for phylogenetic uncertainty, we ran analyses across a
sample of 4,200 of the most likely phylogenetic trees drawn from the posterior
distribution of a Bayesian phylogenetic analysis47. The covariance matrix
representing geographic distance was calculated using the haversine formula
and the longitude and latitude coordinates at the centre of the geographic region
inhabited by each culture (see Supplementary Table 1 for the longitude and
latitude coordinates).
The PGLS-spatial method is capable of simultaneously estimating the effects
of common ancestry, geographic proximity and each of the predictor variables on
the length of conversion time8. To estimate the relative importance of common
ancestry and geographic proximity, we performed an exhaustive grid search in
which the weighting of the geographic and phylogenetic distance matrices ranged
from 0 to 1 in increments of 0.01. This grid included models with a phylogenetic
and spatial effect, a phylogenetic effect without a spatial effect, a spatial effect
without a phylogenetic effect, and neither a spatial nor a phylogenetic effect. To
test the importance of geographic and spatial effects on the length of conversion
time, we used log-likelihood ratio tests to compare the fit of each of these model
structures. All analyses were performed in R (version 3.4.4)56 using a script from
Freckleton and Jetz8, and the packages Ape51 and phytools57.
Reporting Summary. Further information on experimental design is available in
the Nature Research Reporting Summary linked to this article.
Code availability. The analyses used in this paper are based on the statistical
methods developed by Freckleton and Jetz8. Any additional code used in this paper
is available from the first author upon request.
Data availability. The data used in this study are available in Supplementary
Table 1.
Received: 22 October 2017; Accepted: 18 June 2018;
Published: xx xx xxxx
References
1. Johnson, T. M. & Grim, B. J. e World’s Religions in Figures: An Introduction
to International Religious Demography (Wiley, Hoboken, 2013).
2. Deissmann, A. New Light on the New Testament: From Records of the
Graeco-Roman Period (T. & T. Clark, Edinburgh, 1908).
3. Judge, E. A. e early Christians as a scholastic community. J. Relig. Hist. 1,
4–15 (1960).
4. Hopkins, K. Christian number and its implications. J. Early Christ. Stud. 6,
185–226 (1998).
5. Stark, R. e class basis of early Christianity: inferences from a sociological
model. Sociol. Anal. 47, 216–225 (1986).
6. Stark, R. e Rise of Christianity (Princeton Univ. Press, Princeton, NJ, 1997).
7. Watts, J. et al. Pulotu: database of Austronesian supernatural beliefs and
practices. PLoS ONE 10, e0136783 (2015).
8. Freckleton, R. P. & Jetz, W. Space versus phylogeny: disentangling
phylogenetic and spatial signals in comparative data. Proc. R. Soc. B 276,
21–30 (2009).
9. Fox, R. L. Pagans and Christians (Penguin, London, 1986).
10. Barnes, T. D. Constantine and Eusebius (Harvard Univ. Press, Cambridge,
MA, 1981).
11. Nietzsche, F. On the Genealogy of Morality: A Polemic (Hackett Publishing
Company, Indianapolis, IN, 1998).
12. Muthukrishna, M., Shulman, B. W., Vasilescu, V. & Henrich, J. Sociality
inuences cultural complexity. Proc. R. Soc. B 281, 20132511 (2013).
13. Derex, M., Beugin, M. P., Godelle, B. & Raymond, M. Experimental
evidence for the inuence of group size on cultural complexity. Nature 503,
389–391 (2013).
14. Derex, M. & Boyd, R. Partial connectivity increases cultural accumulation
within groups. Proc. Natl Acad. Sci. USA 113, 2982–2987 (2016).
15. Powell, A., Shennan, S. & omas, M. G. Late Pleistocene demography and
the appearance of modern human behavior. Science 324, 1298–1301 (2009).
16. Henrich, J. Demography and cultural evolution: how adaptive cultural
processes can produce maladaptive losses: the Tasmanian case. Soc. Am.
Archaeol. 69, 197–214 (2004).
17. Aoki, K., Lehmann, L. & Feldman, M. W. Rates of cultural change and
patterns of cultural accumulation in stochastic models of social transmission.
eor. Popul. Biol. 79, 192–202 (2011).
18. Rogers, E. M. Diusion of Innovations (e Free Press, New York, NY, 1995).
19. Goodenough, W. H. Oceania and the problem of controls in the study of
cultural and human evolution. J. Polyn. Soc. 66, 146–155 (1957).
20. Kirch, P. V. & Green, R. C. Hawaiki, Ancestral Polynesia: An Essay in
Historical Anthropology (Cambridge Univ. Press, Cambridge, 2001).
21. Murdock, G. P. & Provost, C. Measurement of cultural complexity. Ethnology
12, 379–392 (1973).
22. Sahlins, M. D. Social Stratication in Polynesia (Univ. Washington Press,
Seattle, WA, 1958).
23. Mace, R. & Pagel, M. e comparative method in anthropology. Cur r.
Anthropol. 35, 549–564 (1994).
24. Matthews, L. J., Passmore, S. H., Richard, P. M., Gray, R. D. & Atkinson, Q. D.
Shared cultural history as a predictor of political and economic changes
among Nation States. PLoS ONE 11, e0152979 (2016).
25. Lieber, M. in Exiles and Migrants in Oceania 35–67 (Univ. Press of Hawai’i,
Honolulu, HI, 1977).
26. Keesing, R. Kwaio Religion: e Living and the Dead in a Solomon Island
Society (Columbia Univ. Press, New York, NY, 1982).
27. McLaren, D. et al. Incorporating sociocultural beliefs in mental
health services in Kwaio, Solomon Islands. Australas. Psychiatry 17,
S125–S127 (2009).
28. Gilson, R. P. Samoa 1830–1900: e Politics of a Multi-Cultural Community
(Oxford Univ. Press, Oxford, 1970).
29. Beatty, A. Society and Exchange in Nias (Clarendon Press, Oxford, 1992).
30. Gilson, R. & Crocombe, R. e Cook Islands 1820–1950 (Victoria Univ. Press,
Wellington, 1980).
31. Stark, R. & Bainbridge, W. S. Networks of faith: interpersonal bonds and
recruitment to cults and sects. Am. J. Sociol. 85, 1376–1395 (1980).
NATURE HUMAN BEHAVIOUR | www.nature.com/nathumbehav

Letters Nature HumaN BeHaviour
32. Eerson, C., Lalive, R., Richerson, P. J., McElreath, R. & Lubell, M.
Conformists and mavericks: the empirics of frequency-dependent cultural
transmission. Evol. Hum. Behav. 29, 56–64 (2008).
33. Mesoudi, A. & Lycett, S. J. Random copying, frequency-dependent copying
and culture change. Evol. Hum. Behav. 30, 41–48 (2009).
34. Watts, J. et al. Clarity and causality needed in claims about Big Gods. Behav.
Brain Sci. 39, e27 (2016).
35. Atkinson, Q. Q. D., Latham, A. A. J. & Watts, J. Are Big Gods a big deal in
the emergence of big groups? Relig. Brain Behav. 5, 266–274 (2015).
36. Norenzayan, A. et al. e cultural evolution of prosocial religions. Beh av.
Brain Sci. 39, 1–86 (2014).
37. Diamond, J. Guns, Germs, and Steel: e Fates of Human Societies
(W.W. Norton & Company, New York, NY, 1997).
38. Henrich, J. et al. Understanding cumulative cultural evolution. Proc. Natl
Acad. Sci. USA 113, E6724–E6725 (2016).
39. Kline, M. A. & Boyd, R. Population size predicts technological complexity in
Oceania. Proc. R. Soc. Lond. B 277, 2559–2564 (2012).
40. Collard, M., Ruttle, A., Buchanan, B. & O’Brien, M. J. Population size and
cultural evolution in nonindustrial food-producing societies. PLoS ONE 8,
e72628 (2013).
41. Collard, M., Buchanan, B., O’Brien, M. & Scholnick, J. Risk, mobility or
population size? Drivers of technological richness among contact-period
western North American hunter-gatherers. Phil. Trans. R. Soc. Lond. B 368,
20120412 (2013).
42. Read, D. An interaction model for resource implement complexity based on
risk and number of annual moves. Am. Antiq. 73, 599–625 (2008).
43. Buchanan, B., O’Brien, M. J. & Collard, M. Drivers of technological richness
in prehistoric Texas: an archaeological test of the population size and
environmental risk hypotheses. Archaeol. Anthropol. Sci. 8, 625–634 (2015).
44. Vaesen, K., Collard, M., Cosgrove, R. & Roebroeks, W. Population size does
not explain past changes in cultural complexity. Proc. Natl Acad. Sci. USA
113, E2241–E2247 (2016).
45. Mesoudi, A. Cultural Evolution: How Darwinian eory Can Explain Human
Culture and Synthesize the Social Sciences (Univ. Chicago Press, Chicago,
IL, 2011).
46. Gray, R. D. & Watts, J. Cultural macroevolution matters. Proc. Natl Acad. Sci.
USA 114, 7846–7852 (2017).
47. Gray, R. D., Drummond, A. J. & Greenhill, S. J. Language phylogenies reveal
expansion pulses and pauses in Pacic settlement. Science 323, 479–483 (2009).
48. Sheehan, O., Watts, J., Gray, R. D. & Atkinson, Q. D. Coevolution of
landesque capital intensive agriculture and sociopolitical hierarchy. Proc. Natl
Acad. Sci. USA 115, 3628–3633 (2018).
49. Blomberg, S. P., Garland, T. & Ives, A. Testing for phylogenetic signal in
comparative data: behavioral traits are more labile. Evolution 57,
717–745 (2003).
50. Kembel, S. W. et al. Picante: R tools for integrating phylogenies and ecology.
Bioinformatics 26, 1463–1464 (2010).
51. Paradis, E., Claude, J. & Strimmer, K. APE: analyses of phylogenetics and
evolution in R language. Bioinformatics 20, 289–290 (2004).
52. Dow, M. & E, E. Global, regional, and local network autocorrelation
in the standard cross-cultural sample. Cross Cult. Res. 42,
148–171 (2008).
53. Atkinson, Q. D., Coomber, T., Passmore, S. H., Greenhill, S. J. & Kushnick, G.
Cultural and environmental predictors of pre-European deforestation on
Pacic Islands. PLoS ONE 11, e0156340 (2016).
54. Sookias, R. B., Passmore, S. H. & Atkinson, Q. D. Deep cultural ancestry and
human development indicators across nation states. R. Soc. Open Sci. 5,
171411 (2018).
55. Greenhill, S. J., Blust, R. & Gray, R. D. e Austronesian Basic
Vocabulary Database: from bioinformatics to lexomics. Evol. Bioinform. 4,
271–283 (2008).
56. R Core Development Team. R: A Language and Environment for Statistical
Computing (R Foundation for Statistical Computing, 2015).
57. Revell, L. J. phytools: an R package for phylogenetic comparative biology (and
other things). Methods Ecol. Evol. 3, 217–223 (2012).
Acknowledgements
The authors thank S. Passmore for assistance with the PGLS-spatial analyses, and
A. Powell for comments. They also thank the John Templeton Foundation (28745),
Templeton Religion Trust (TRT0153) and Marsden Fund (UOA1104) for funding. The
funders had no role in study design, data collection and analysis, decision to publish, or
preparation of the manuscript.
Author contributions
J.W. and Q.D.A. designed the study. J.W. and O.S. developed the coding scheme and coded
the data. J.W. ran the analyses. J.W. wrote the manuscript with input from all authors.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information is available for this paper at https://doi.org/10.1038/
s41562-018-0379-3.
Reprints and permissions information is available at www.nature.com/reprints.
Correspondence and requests for materials should be addressed to J.W.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
published maps and institutional affiliations.
NATURE HUMAN BEHAVIOUR | www.nature.com/nathumbehav

1
nature research | reporting summary April 2018
Corresponding author(s): Joseph Watts
Reporting Summary
Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Research policies, see Authors & Referees and the Editorial Policy Checklist.
Statistical parameters
When statistical analyses are reported, confirm that the following items are present in the relevant location (e.g. figure legend, table legend, main
text, or Methods section).
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
An indication of whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested
A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistics including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND
variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)
For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.
For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r), indicating how they were calculated
Clearly defined error bars
State explicitly what error bars represent (e.g. SD, SE, CI)
Our web collection on statistics for biologists may be useful.
Software and code
Policy information about availability of computer code
Data collection No specific software was used in data collection.
Data analysis We use R with an additional script from Freckleton, R. P., & Jetz, W. (2009). Space versus phylogeny: disentangling phylogenetic and
spatial signals in comparative data. Proceedings of the Royal Society of London B: Biological Sciences, 276(1654), 21-30.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors/reviewers
upon request. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.
Data
Policy information about availability of data
All manuscripts must include a data availability statement. This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A list of figures that have associated raw data
- A description of any restrictions on data availability
The data used in this study is available in the supplementary data file.

2
nature research | reporting summary April 2018
Field-specific reporting
Please select the best fit for your research. If you are not sure, read the appropriate sections before making your selection.
Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/authors/policies/ReportingSummary-flat.pdf
Behavioural & social sciences study design
All studies must disclose on these points even when the disclosure is negative.
Study description This study uses a quantitative cross-cultural research method.
Research sample Historical, Austronesian speaking societies.
Sampling strategy All societies from the Pulotu database that have sufficient historical data.
Data collection Data were gathered from the records of missionaries, explorers, and ethnographers, as well as official government censuses
(Supplementary Table 1). Data collectors were not blind to the hypotheses of the study. The predictor variables were collected as part of
a broader database (Pulotu) and were not specifically designed for this study.
Timing Data was collected and coded between 2015-2017
Data exclusions We coded all societies from the Pulotu database for which we found sufficient data. The Pulotu database contains 96 societies that can
be linked to an established language-based phylogeny, which is required for the use of the PGLS-spatial method. Of these 96 societies, 26
societies were excluded due to a lack of sufficient historical data on the length of conversion times to Christianity, or because they
converted to a different major world religion, such as Islam. These exclusion criteria were decided in advance of the data analysis.
Non-participation Not applicable, this is not a participant based study.
Randomization Not applicable, this is not an experimental study.
Reporting for specific materials, systems and methods
Materials & experimental systems
n/a Involved in the study
Unique biological materials
Antibodies
Eukaryotic cell lines
Palaeontology
Animals and other organisms
Human research participants
Methods
n/a Involved in the study
ChIP-seq
Flow cytometry
MRI-based neuroimaging