Pathogen Prevalence, Group Bias, and Collectivism in the Standard Cross-Cultural Sample

Abstract

It has been argued that people in areas with high pathogen loads will be more likely to avoid outsiders, to be biased in favor of in-groups, and to hold collectivist and conformist values. Cross-national studies have supported these predictions. In this paper we provide new pathogen codes for the 186 cultures of the Standard Cross-Cultural Sample and use them, together with existing pathogen and ethnographic data, to try to replicate these cross-national findings. In support of the theory, we found that cultures in high pathogen areas were more likely to socialize children toward collectivist values (obedience rather than self-reliance). There was some evidence that pathogens were associated with reduced adult dispersal. However, we found no evidence of an association between pathogens and our measures of group bias (in-group loyalty and xenophobia) or intergroup contact.

Full text

Pathogen Prevalence, Group Bias, and Collectivism
in the Standard Cross-Cultural Sample
Elizabeth Cashdan &Matthew Steele
Published online: 7 February 2013
#Springer Science+Business Media New York 2013
Abstract It has been argued that people in areas with high pathogen loads will be
more likely to avoid outsiders, to be biased in favor of in-groups, and to hold
collectivist and conformist values. Cross-national studies have supported these pre-
dictions. In this paper we provide new pathogen codes for the 186 cultures of the
Standard Cross-Cultural Sample and use them, together with existing pathogen and
ethnographic data, to try to replicate these cross-national findings. In support of the
theory, we found that cultures in high pathogen areas were more likely to socialize
children toward collectivist values (obedience rather than self-reliance). There was
some evidence that pathogens were associated with reduced adult dispersal. However,
we found no evidence of an association between pathogens and our measures of
group bias (in-group loyalty and xenophobia) or intergroup contact.
Keywords Infectious disease .Historical pathogen prevalence .In-group bias .
Collectivism .Cross-cultural analysis
After the September 11, 2001, terrorist attacks, three-quarters of Americans responded
by displaying American flags (Skitka 2005). Threat to the group is a potent factor in
promoting in-group loyalty, as has been shown both experimentally and ethnographi-
cally. In an earlier cross-cultural study using the standard cross-cultural sample (SCCS),
we found loyalty to the larger society to be associated with threats from catastrophic
famine as well as from external warfare and interethnic violence (Cashdan 2001). It is
plausible that affiliation with an in-group can effectively counter many types of threats,
through mutual help and support, and that identifying and sharpening intergroup
boundaries is a necessary feature of this process (e.g., Brewer 1999).
Hum Nat (2013) 24:59–75
DOI 10.1007/s12110-012-9159-3
Electronic supplementary material The online version of this article (doi:10.1007/s12110-012-9159-3)
contains supplementary material, which is available to authorized users.
E. Cashdan (*):M. Steele
Department of Anthropology, University of Utah, Salt Lake City, UT 84112, USA
e-mail: cashdan@anthro.utah.edu
Author's personal copy

Recently, attention has turned to the role of another type of threat, infectious
disease, in promoting collectivism and group bias. Social and behavioral responses
to disease should be especially important in traditional anthropological populations of
the twentieth century, where modern medicine was often scarce. In this study,
therefore, we extend our previous work on ethnicity and threat in the standard
cross-cultural sample by focusing on the threat of pathogen exposure and by broad-
ening the group bias measures to consider collectivist values and intergroup mobility.
The Pathogen Theory of Group Bias
Recent studies of the “behavioral immune system”(Schaller and Duncan 2007;
Schaller and Murray 2010) have argued that ethnocentrism and xenophobia are
heightened where disease is prevalent or salient because they lead to avoidance of
outsiders and the novel pathogens those outsiders carry. Fincher and Thornhill
(2008a,b) argue that we evolve resistance to local pathogenic strains, which makes
novel pathogens—those carried by outsiders—relatively more dangerous. Under
conditions of high pathogen prevalence, therefore, avoidance of outsiders should be
favored.
Some of the most compelling evidence in support of this theory comes from
experimental studies showing that people who feel more vulnerable to disease (either
as a baseline trait or through experimental manipulation) are more xenophobic toward
immigrants (Faulkner et al. 2004) and more ethnocentric (Navarrete and Fessler
2006). Cross-nationally, greater pathogen prevalence has been associated with a
variety of traits that might lead to avoidance of outsiders, including xenophobia
(Faulkner et al. 2004) and lower levels of extraversion and openness to experience
(Schaller and Murray 2008; Thornhill et al. 2010). Pathogen prevalence has also been
associated with collectivist values (Fincher et al. 2008; Thornhill et al. 2010), and
family loyalty and religiosity (Thornhill et al. 2010). We refer to these ideas collec-
tively as the pathogen theory of group bias.
We know little as yet about whether this theory explains variation in the largely
small-scale societies of the SCCS, although we would expect infectious disease to be
salient in such societies. Fincher and Thornhill (2008a) have used cross-cultural data
to argue that “high contagion risk associated with out-group contact”is associated
with limited dispersal in hunter-gatherers. They demonstrate this relationship by
inferring pathogen contagion risk from latitude, and out-group contact from
Binford’s(2001) forager range-size data. They show that latitude and range size are
highly correlated in hunter-gatherers, even when controlling for population size and
dependence on hunting. The interpretation warrants caution since range size is
determined by the distribution of foraging resources (plant distributions and animal
ranges), and resource distributions are strongly shaped by latitude. It is also not clear
that intergroup contact among foragers increases with greater range size, since the
latter is a function of foraging strategies. It seems plausible, therefore, that the small
ranges of low-latitude foragers are shaped directly by the distribution of their food
resources (as discussed by Binford and others) rather than by avoidance of exotic
pathogens. Additional tests of this relationship would be helpful, so we look at
mobility variables that measure intergroup contact more directly.
60 Hum Nat (2013) 24:59–75
Author's personal copy

Collectivism
Avoidance of outgroups should reduce contact with novel pathogens, but it is less
obvious that in-group loyalty and related collectivist values would do the same. Yet
the cross-national studies cited above found that pathogen risk is associated with
countries holding more collectivist, as opposed to individualistic, values, and with
related traits such as family loyalty and religiosity. Fincher and Thornhill (2012)
suggest that this association stems from the benefits of collectivism in providing
support and care during times of illness, an argument that would be applicable to
other threats as well. Gelfand et al. (2011) have used a related construct, societal
“tightness”and “looseness,”to show that “tight”societies (those with strong norms
and sanctions) are also more likely to be found where environmental threats, includ-
ing infectious disease, are more severe. They explain this relationship by suggesting
that such societies benefit from values that enhance order and social coordination so
as to deal effectively with environmental threats. We therefore included two measures
that might reflect these societal values: pressure on children toward obedience vs.
self-reliance and loyalty to one’s community and society. We want to know how these
traits are associated with each other and with intergroup contact, and whether they are
related to pathogen prevalence.
In this paper, therefore, we aim to see whether the threat of pathogens is associated
with collectivist values, group bias, and intergroup contact in the SCCS, and whether
these dependent variables are associated with each other. If xenophobia and a
reduction in intergroup contact are associated with pathogens, it would be consistent
with the hypothesis that these behaviors serve to prevent exposure to novel patho-
gens. If ethnocentrism and socialization toward obedience are associated with patho-
gens, it would be more consistent with the general threat-reducing hypotheses
reviewed above (mutual support and/or better coordination).
Methods
Sample
The Standard Cross-Cultural Sample (Murdock and White 1969) consists of 186
nonindustrial, mostly small-scale cultures. The sample is stratified by language and
region to minimize similarities from shared history and culture. We use it for that
reason and because it has been the basis for many independently coded variables of
relevance to this research question. All variables, other than our new pathogen codes
described below, were taken from these published sources. The codes have been
republished and in some cases updated in the electronic journal World Cultures.We
used the 2003 data disks from that journal for analysis, and the original sources, as
cited below, for full variable definitions, coding procedures, and related information.
New Pathogen Codes
We developed new pathogen codes for the SCCS using historical sources, chiefly
global maps created between the 1930s and 1950s. We coded eight pathogens:
Hum Nat (2013) 24:59–75 61
Author's personal copy

leishmanias, trypanosomes, malaria, schistosomes, filariae, dengue, typhus, and
plague. We followed the sources and methods used by Murray and Schaller (2010)
in their historical cross-national pathogen codes but made the codings specific to the
local conditions (within 200 km) surrounding the focal SCCS societies. Our main
sources were the three-volume series of maps in Rodenwaldt and Bader (1961) and
the maps and data in Simmons et al. (1944). Where these sources disagreed, we gave
priority to Rodenwaldt and Bader. For two pathogens we also referred to information
in Faust and Russell (1964).
We used historical sources on infectious disease because most of the SCCS
societies have focal dates in the mid twentieth century, and many pathogens, and
the vectors that transmit them, have become worldwide travelers since that time.
Maps of infectious disease prevalence have changed markedly over the past decades
and centuries owing to conquest, inter-regional travel, environmental changes, and
advances in public health. Our data on disease prevalence are reasonably closely
matched in time to the target dates of most of the SCCS societies. A few societies had
very early focal dates, with information coming from historical rather than ethno-
graphic sources. Removing them did not materially change the results, so they are
included in these analyses.
Low (1990,1994) has published an excellent set of pathogen codes for the
SCCS, and we initially developed new codes because we hoped to get a dataset
diverse and fine-grained enough to test hypotheses concerning mode of trans-
mission and zoonotic vs. non-zoonotic diseases. We were largely unsuccessful
owing to limitations of the source material. We use our code here, therefore,
not because we think it is better than Low’s, but to provide an independent
check, with a somewhat different set of diseases, on our conclusions. Five
pathogens were in both sets, three were only in ours, two were only in
Low’s. We therefore also developed a combined code to take advantage of all
ten pathogens. We ran all analyses using all three datasets, and we report if the
datasets disagree. Unless otherwise noted, statistics refer to the combined code
(explained below).
We coded disease prevalence for each pathogen on a four-point scale. The ESM
lists the values for two pathogen indices. The first, the pathogen sum score, is the sum
of the scores (1–4) of each pathogen we coded. Most of the pathogens are really
clusters of related species, but in most cases we had only aggregate data. Where we
had separate data for different forms, as we did for schistosomes and leishmanias, the
sum score was based on the highest value. For example, if a location had a “4”for one
form of leishmanias and a “2”for another form, we gave it a score of 4 for that
pathogen.
The second index, the pathogen z-score, is more complicated but contains more
data. First, it adds to our data the scores for two pathogens (leprosy and spirochetes)
that Low coded but we did not. Because the two scales were different, we normalized
the scores for all pathogens and used the resulting z-scores rather than the raw values.
The normalization procedure also allowed us to make finer distinctions for our
codings of schistosomes and leishmanias. In this case, if a location had a “4”for
one form and a “2”for another form, we gave the pathogen a value of 6 and used the
z-score for that pathogen when calculating its prevalence. The pathogen z-score
index, therefore, is the sum of these z-scores.
62 Hum Nat (2013) 24:59–75
Author's personal copy

Biases in our pathogen data (and Low’s) have potential consequences for our
results, and for any analyses that use either dataset. As Table 1shows, the datasets are
weighted toward pathogens that are transmitted through arthropod and other vectors
(i.e., leishmanias, trypanosomes, schistosomes, malaria, filariae, typhus, and dengue).
A few of these are also zoonotic; that is, they have significant non-human reservoirs.
The prevalence of such diseases is strongly shaped by the geographic distribution of
the vectors that transmit them and the species that host them. It is well known that the
disease burden is greater in the tropics, but this bias toward diseases transmitted
through vectors probably makes the correlation between disease and latitude even
stronger than it would be with a more inclusive set of pathogens. Both datasets,
therefore, underrepresent diseases spread through human-to-human contact, includ-
ing diseases spread by droplet infection (for example, tuberculosis and measles) and
through fecal-oral transmission (for example, cholera and typhoid). These diseases
are historically important, but they are excluded from our data (and probably other
historical datasets) because we could not find adequate worldwide historical data on
their prevalence. The implications of this bias are addressed in later sections.
Dependent Variables: Collectivist Values, Group Bias, and Inter-group Mobility
We used previously coded data for the SCCS to measure collectivist values, in-group
bias, residential mobility, and contact with other groups. Variables were reverse-
coded where necessary to ensure that large numbers always indicate “more”of
something (greater mobility, more contact, more ethnocentrism, etc.).
Table 1 Pathogens coded, and mode of transmission
Pathogen Coder Transmission mode
1 Leprosy L Human–human: droplet
2 Spirochetes L
Syphilis Human–human: venereal
Bejal, pinta, yaws Human–human: skin contact
Relapsing fever, lyme Vector: tick, lice
3 Dengue C-S Vector: mosquito
4 Typhus C-S Vector: lice, flea, ticks, mites
5 Plague C-S
Bubonic Vector: flea
Pneumonic Human–human: droplet
6 Filariae L, C-S
Guinea worm Vector: copepod (aquatic crustacean)
Lymphatic, onchocerciasis Vector: mosquito, blackfly
7 Schistosomes L, C-S Vector: aquatic snail
8 Leishmanias L, C-S Vector: sandfly
9 Trypanosomes L, C-S Vector: tsetse fly, tiratomid bug
10 Malaria L, C-S Vector: mosquito
LLow, C-S Cashdan and Steele
Hum Nat (2013) 24:59–75 63
Author's personal copy

Obedience and Self-reliance
Collectivist societies are distinguished from individualistic ones in valuing the group
over the individual. We operationalized this set of values by using codes that
measured the degree to which children were socialized toward obedience
(collectivism) or self-reliance (individualism). The data are a subset of the childhood
socialization codes of Barry et al. (1976). Self-reliance indicates initiative and
encouragement of children to act without supervision; obedience indicates pressure
to obey requests by parents and others in authority.
Barry et al. (1976) published separate codes for young girls, young boys, older
girls, and older boys and measured the strength of socialization pressure toward these
values on a scale of 0 (none) through 12 (extremely strong). There are interesting sex
differences in these socialization patterns (Low 1989), but for the purposes of this
study we summed the values for girls and boys of both age groups in order to create a
global measure of the degree to which these values were emphasized in the culture.
Intergroup Mobility and Group Bias
Political codes developed by Ross (1983) were used for adult dispersal, intergroup
contact, local and societal loyalty, and hostility to outside groups. These variables
were coded for alternate societies in the SCCS. Even-numbered societies were
selected unless available data were inadequate, in which case an adjacent society
was substituted. Societies with adjacent numbers are most closely related linguisti-
cally, so coding alternate societies enhances independence.
Adult dispersal is an ordinal scale that measures whether adults generally remain
with the same community throughout their lives, occasionally move to new commu-
nities, or typically move to new communities. Intergroup contact is an ordinal scale
that measures whether contact with persons from other societies is frequent, occurs
occasionally, or is rare or nonexistent. The type of contact is described as including
“trade, warfare, travel, etc.”The two measures differ in scale as well as type of
mobility, with the dispersal measure referring to residential movement between
communities (presumably of the same society) and the contact measure referring to
non-residential mobility that fosters contact with people from other societies.
The pathogen theory also predicts that high pathogen risk will enhance group bias
(in-group preference and out-group prejudice). Ross (1983:180) coded group bias at
the level of both the local community and the broader society. At the local level, the
variable measures the degree of “in-group loyalty, or we feeling, directed towards the
local community,”while at the societal level it measures “in-group loyalty, or we
feeling, directed towards the wider society.”Both variables were measured on an
ordinal scale (especially high, high, moderate, low). We refer to in-group loyalty at
the level of the wider society as “ethnocentrism”for convenience, even though it is
possible to have a feeling of pride in one’s group without necessarily thinking it
superior to other societies.
Out-group hostility is defined as “hostilitytowardothersocieties....thisvariable
seeks to evaluate the feelings towards other societies.”It is measured on a four-point
ordinal scale from “low hostility is shown towards outsiders”to “hostility is extensive—
quite bitter feelings exist towards almost all outsiders.”We use the word “xenophobia”
64 Hum Nat (2013) 24:59–75
Author's personal copy

interchangeably with “out-group hostility”or “hostility to other societies”when refer-
ring to this variable. Because hostility toward outsiders is restricted here to members of
other societies, it is similar in scale to the wider of the two ethnic loyalty measures,
which refers to loyalty to the wider society.
Other Variables
We used latitude (Murdock and White 1969) and population density (Murdock and
Wilson 1972) as control variables in our correlations. Population density is a seven-
point ordinal scale that measures the “density of population in area exploited or
controlled by the focal or typical community.”It ranges from 1 (less than 1 person/5
sq. mi) to 7 (more than 500 persons/sq. mi).
We also used several additional variables in selected analyses. Levels of political
integration (Murdock and Provost 1971) is a five-item scale from 0 (no political
authority even at the community level) through 4 (states with three or more admin-
istrative levels above the local level). Community size is an eight-item scale that
measures the size of the focal community (Murdock and Wilson 1972). Number of
frost-free months and mean annual temperature (measured from the nearest weather
stations to the focal society) were coded by Whiting (1989).
Analysis
Sample sizes varied across the different variables, and the missing values differ from
variable to variable. An analysis that used listwise deletion (limiting analyses to only those
cases without missing data on any variable) would end up with next to no data, so we used
pairwise deletion, calculating correlations on all data available for each pair of variables.
SAS was used for all analyses. When the dependent variable was ordinal,
Spearman’s rank order correlations were used in the analysis. Because latitude and
population density are strongly correlated with pathogen prevalence, we also calcu-
lated partial correlations to control for these variables. SAS calculates partial corre-
lations for Spearman’s correlations in the same way it does for Pearson correlations,
and the result can be interpreted similarly. The algorithm for calculating the correla-
tions and the associated probability values is given in the SAS documentation (SAS
Institute 2012; see also the discussion in section 3.9 of Shipley 2002). Ordinal logistic
regression would have been appropriate for the mobility and group bias variables
since they had a small number of ordered responses, but the sample size was too small
to give sufficient power (these variables were coded for only half the societies in the
SCCS). We also used multiple regression to explain (with caveats) the patterning in
the obedience and self-reliance data.
The Standard Cross-Cultural Sample was chosen to maximize independence of the
cases; as noted above, it is stratified by both language and region for this purpose.
Nonetheless, concerns have been raised about whether the cases are truly indepen-
dent, and various corrections have been proposed, both phylogenetic (using language
as a proxy for similarities owing to cultural evolution) and spatial. We used latitude as
a control variable in our correlations because of its association with disease preva-
lence, but doing so has the added advantage of providing additional protection against
cultural similarities that arise from similar environments and spatial proximity.
Hum Nat (2013) 24:59–75 65
Author's personal copy

Results
Pathogen Data and Distributions
Cronbach’s alpha for the eight-item pathogen sum score was 0.76, and for the ten-
item pathogen z-score was 0.81. Five of the pathogens we coded were also coded by
Low: leishmanias, trypanosomes, malaria, schistosomes, and filariae (Table 1).
Summing pathogen scores for just these five, we find good agreement between our
codes: r=0.80, n=186, p<0.0001 (correlations between the scores of the individual
pathogens varied between r
s
=0.61 and r
s
=0.75). In addition to these five, Low coded
spirochetes and leprosy, while we coded dengue, typhus, and plague. Correlations
between Low’s overall score and ours was also high (r=0.80, n=186, p<0.0001).
Pathogens are correlated with latitude in both datasets. The correlation of the
pathogen z-score index with latitude (absolute value) was r
s
=−0.53, p<0.0001, n=
186; the correlation with Low’s pathogen code was r
s
=−0.65, p<0.0001, n=186. As
noted in the discussion of methods, this probably reflects not just the well-known
heavy infectious disease burden in the tropics, but biases in both datasets toward
zoonotic diseases and diseases that are transmitted through insect and other vectors.
Within the tropics themselves, we found no relationship between pathogen preva-
lence and distance from the equator (r
s
=−0.05, p=0.6, n=111), whereas away from
the tropics the association was strong (r
s
=−0.72, p<0.0001, n=75).
Pathogen loads are low in high-latitude areas, although within this region the
societies in our dataset with a frost-free climate (chiefly coastal areas) have higher
pathogen loads. Mean annual temperature does not affect pathogen prevalence when
latitude is controlled, but frost-free climate does: the correlation between number of
frost-free months (Whiting 1989) and pathogen prevalence, controlling for latitude, is
r
s
=0.52, p=0.001, for n=38 societies at or above 40°.
Population density is an equally important predictor of pathogen load. The relation-
ship between population density (Murdock and Wilson 1972) and pathogen prevalence
is r
s
=0.46, p<0.0001, n=184, and is stronger in non-tropical regions (r
s
=0.69, p<
0.0001, n=75) than tropical ones (r
s
=0.17, p<0.07, n=109). Partial correlations be-
tween pathogens and density controlling for latitude show similar results.
Because both latitude and population density are correlated with many variables
other than pathogens, some of which are likely to have independent effects on
mobility, ethnocentrism, and xenophobia, we control for latitude and population
density in our pathogen analyses.
Collectivism in the SCCS
Socialization for self-reliance, which we interpret as a measure of individualistic
values, was negatively correlated with socialization for obedience (r
s
=−0.23, p=
0.007, n=135). As Table 2shows, the social variables (community size, density,
political complexity) that positively predicted obedience also negatively predicted
self-reliance. Taken together, these results support our initial assumption that
obedience and self-reliance are related variables associated with two ends of the
same dimension. In addition to analyzing the two variables separately, therefore,
we subtracted self-reliance from obedience to create an index, which we view as
66 Hum Nat (2013) 24:59–75
Author's personal copy

a proxy for socialization toward collectivist values. We refer to it as the collec-
tivism index.
More-collectivist societies, by this measure, are larger (as measured by community
size), denser, and more politically complex (Table 2). Political complexity—defined
as levels of political integration (Murdock and Provost 1971)—was the strongest
predictor. As Low (1989) and others have noted, large, complex societies place a
premium on traits that foster group harmony at the expense of individual autonomy.
We had expected these socialization values to be associated with the other dependent
variables under study, but they were not (Table 2): societies that socialized for obedience
did not have greater feelings of in-group loyalty or out-group hostility; nor did societies
that socialized for self-reliance have fewer such feelings. Obedience and self-reliance
were also not associated with adult dispersal or inter-group contact. Sample sizes were
low for these variables (they were coded for only half the SCCS sample), but the lack of
consistent trends suggests that this is not just a result of low power.
The most interesting relationship among our dependent variables was between group
bias and intergroup mobility, and it ran counter to our expectations. Whereas contact
with outside groups might be thought to foster (or reflect) a reduction in group bias, the
data indicated the opposite: the greater the contact with other societies, the greater the
hostility toward them (r
s
=0.25, p=0.04, n=69) and the greater the loyalty to one’sown
society (r
s
=0.32, p=0.006, n=74). Similarly, loyalty to one’s local community was
greater where adult dispersal was more prevalent (r
s
=0.31, p=0.01, n= 68). These
patterns gain strength from the fact that the correlations are found only at the equivalent
scales: between communities for adult dispersal and community loyalty, and between
societies for intergroup contact, xenophobia, and loyalty to the wider society.
Patterning within the group bias variables themselves was inconsistent: As de-
scribed in Cashdan (2001), hostility to other societies (xenophobia) was not
Table 2 Predictors of socializa-
tion for obedience and self-
reliance: Spearman correlations
(sample sizes in parentheses)
*p<.05, **p< .01, ***p< .001,
****p0.0001
Obedience Self-reliance
Independent variables
Pathogens .16* (158) −.20* (149)
Community size .19* (158) −.17* (149)
Density .24** (157) −.17* (147)
Political complexity .37*** (158) −.22** (149)
Partialling latitude
Pathogens .36**** (158) −.24** (149)
Community size .21** (158) −.17* (149)
Density .31**** (157) −.17* (147)
Political complexity .37**** (158) −.22** (149)
Dependent variables
Community loyalty .01 (66) .08 (65)
Society loyalty .00 (67) .13 (66)
Xenophobia .09 (60) −.10 (58)
Dispersal −.08 (61) .13 (60)
Intergroup contact −.21 (67) .09 (64)
Hum Nat (2013) 24:59–75 67
Author's personal copy

associated with loyalty to one’s own society (ethnocentrism). However, loyalty to the
local community did pattern as expected, being associated with both ethnocentrism
(r
s
=0.44, p<0.0001, n=76) and xenophobia (r
s
=0.27, p=0.02, n=68).
The Pathogen Theory of Group Bias: Obedience and Self-reliance
The data support the predicted association between pathogen loads and collectivist values,
with obedience being socialized more strongly and self-reliance less strongly in high-
pathogen areas. The pattern is stronger when latitude is partialled out, and it remains when
population density is also partialled out. The partial correlation between obedience and
pathogens, controlling for latitude, was r
s
=0.36, p< 0.0001, n=158, and with self-reliance
was r
s
=−0.24, p=0.004, n=149. When population density was also partialled out, the
pattern remained significant: for obediance r
s
=0.27, p=0.0007, n=157; for self-reliance,
r
s
=−0.17, p=0.04, n=147. These data refer to the combined pathogen z-score; the
relationship with obedience was similar for the different pathogen codes, but with self-
reliance it was non-significant with Low’scode.
We explored these relationships further with a multiple regression, using the
collectivism index (obedience minus self-reliance) as the dependent variable and
pathogen z-score, population density, and levels of political integration as dependent
variables. Two societies with very high collectivist values (Aztec and Inca) were
removed for this analysis. The residuals were normally distributed in all models
tested but some of the variables were not, so we present these results to indicate the
nature of the patterning only. The best fit, with an adjusted R
2
=0.24, was obtained
with latitude, pathogens, and political integration in the model, all of which were
highly significant. Figure 1shows a partial residual plot of collectivism by pathogen
prevalence, controlling for (a) latitude and (b) latitude and political integration. The
Spearman’s correlation between pathogen prevalence and the collectivism index,
partialling latitude and political integration, was r
s
=0.30, p=0.0005, n=134.
Table 3shows bivariate correlations between the variables used in the model.
The Pathogen Theory of Group Bias: Intergroup Mobility
The pathogen theory of group bias predicts that people will minimize adult dispersal
from the natal community and will avoid contact with outsiders as a way to avoid
exposure to pathogens to which they are not well-adapted. We found some support for
this hypothesis. High pathogen prevalence was associated with lower dispersal within
the tropics (r
s
=−0.46, p=0.01, n=28) and, when controlling for latitude, across the
entire sample (r
s
=−0.34, p=0.004, n= 71). With both latitude and population density
controlled, the relationship was still significant but weaker: r
s
=−0.24, p=0.05, n=71.
These statistics refer to the pathogen z-score. The relationships are stronger with our
pathogen sum score and weaker (non-significant trends) with Low’s pathogen index.
Intergroup contact is another measure that might be expected to increase exposure
to exotic pathogens. This variable measures contact between societies and can be due
to a variety of factors, including trade, warfare, and travel. We found no relationship
between this measure of group contact and pathogens (r
s
=−0.09, p=0.43, n=76,
partialling latitude and density), and in tropical regions the trend was in the opposite
direction to that predicted (r
s
=0.27, p=0.07, n=47).
68 Hum Nat (2013) 24:59–75
Author's personal copy

The Pathogen Theory of Group Bias: Xenophobia and Ethnic Loyalty
Pathogens have been hypothesized to foster hostility to other ethnic groups
(xenophobia) because such attitudes would cause people to avoid outsiders and
consequently reduce exposure to novel strains of disease. In an earlier preliminary
analysis using Low’s data, we found that xenophobia was greater in high-pathogen
areas, controlling for latitude (Cashdan 2012). We replicated that result for this study
and found that it holds when population density is also controlled (r
s
=0.23, p=0.05,
Fig. 1 Pathogen prevalence by
socialization for collectivism,
partial residual plots. apartialing
latitude only. bpartialing latitude
and levels of political integration
Hum Nat (2013) 24:59–75 69
Author's personal copy

n=69). However, the relationship does not exist with our new pathogen codes. Our
codes and Low’s differed in the diseases measured, so we looked individually at the
different diseases to see whether we could identify the source of the discrepancy. The
patterning with xenophobia was found with leishmanias (coded in both datasets, and
associated with xenophobia in both of them) and spirochetes (coded only by Low),
but was not significantly associated with other pathogens.
Pathogens have also been hypothesized to foster in-group loyalty, perhaps because
in-group members provide support during illness. Ross (1983) coded loyalty at both
the ethnic group level (ethnocentrism) and the local community level. We expected
that a relationship with pathogens would be most likely with the latter, since direct
support during illness is typically based on one’s personal network of kin and friends.
However, we found no relationship between pathogens and loyalty at either commu-
nity or ethnic-group level, with either pathogen dataset.
Summary and Discussion
The Pathogen Codes: Characteristics and Environmental Determinants
Cross-national data have found associations between greater pathogen prevalence and
a variety of traits reflective of in-group affiliation and avoidance of out-groups. We
created new historical pathogen codes, roughly contemporaneous with the focal dates
of most societies in the standard cross-cultural sample (SCCS), in order to evaluate
those relationships in that sample. Most of the pathogen sources used by Low (1994)
are also historical, and we used both datasets in our analyses.
Because of limitations in the historical data, both our data and Low’s are biased
toward diseases that are transmitted through arthropod and other vectors, some of
which are also zoonotic, i.e., have significant non-human reservoirs. As noted above,
the datasets underrepresent diseases spread through human contact, including those
spread through droplet infection (for example, tuberculosis and measles) and through
fecal-oral transmission (for example, cholera and typhoid). This reduces the problem
of contemporaneity since diseases spread through the latter means are more likely to
have undergone rapid changes in distribution owing to inter-regional mobility and
improvements in public health. However it also introduces a probable bias in
Table 3 Bivariate correlations for variables used in the pathogen model
Latitude Density Political complexity Collectivism
Pathogens −.53*** .46*** .34*** .25**
Latitude −.25*** .02 .18*
Density .59*** .32***
Political complexity .41***
Spearman’s rank-order correlations: n=186 for pathogens, latitude (absolute value), and political complex-
ity (levels of political integration); n=184 for population density; n= 135 for collectivism (socialization for
obedience minus self-reliance).
*p<.05, **p< .01, ***p< .001
70 Hum Nat (2013) 24:59–75
Author's personal copy

geographical distributions since diseases spread through human contact can poten-
tially be found anywhere, whereas diseases spread through vectors or sustained
through animal reservoirs are constrained geographically by the climates in which
the animal vectors and reservoirs can survive.
Distance from the equator was strongly correlated with pathogen prevalence overall,
although not within the tropics themselves. Within high-latitude zones, the presence of
frost had a secondary effect in inhibiting pathogen prevalence, independent of mean
temperature and latitude. We speculate that frost may pose a barrier to arthropod vectors
over and above the other climatological correlates of latitude and thus may be significant
in predicting the disease burden in such areas. The other strong predictor of disease
prevalence was high population density, especially away from the tropics.
Pathogens and Group Boundedness in the SCCS
Previous studies, including cross-national studies, have found that societies with a
high pathogen prevalence hold more collectivist values, and our clearest results were
with this measure. The collectivist-individualist continuum reflects the importance
placed on the social group as opposed to the individual. We used as a proxy the
degree to which children are socialized toward obedience (collectivism) or self-
reliance (individualism). These variables are correlated with levels of political inte-
gation and with population density.
As predicted, areas with higher pathogen prevalence showed more collectivism by this
measure, both greater pressure toward obedience and less pressure toward self-reliance.
Controlling for latitude strengthened the relationship, and it remained significant when
we also controlled for population density and levels of political integration.
There was weak support for the prediction that high pathogen prevalence would
lead to behavior that minimized exposure to outsiders: in areas with high pathogen
loads, adults were less likely to move to other communities (controlling for latitude
and population density). This measure of residential mobility between communities
complements the results of Fincher and Thornhill (2008a) on forager range size.
Contact with people from other societies, on the other hand, showed no such
patterning. We also found no relationship between pathogens and in-group loyalty,
at the level of either the wider society or the local community. Our previously
reported support for a link with xenophobia (Cashdan 2012) was based on Low’s
pathogen codes and was not found in ours.
There were few associations among the dependent variables, and the clearest and
most interesting ran counter to our expectations. We expected contact with outsiders
to reduce group bias, but we found the opposite pattern: contact with people from
other societies was associated with both greater xenophobia and more ethnocentrism.
This could reflect the benign sharpening of boundaries where ethnic groups form
economically interdependent polities (e.g., Barth 1969), but it could also reflect the
reported corrosive effect of ethnic diversity on social capital (Putnam 2007) and
cross-race friendships (Moody 2001). It could also be an artifact of the way inter-
group contact was measured: the variable includes contact from warfare as well as
trade and visiting, and warfare is associated with both xenophobia and ethnocentrism
in the SCCS (Cashdan 2001). The same pattern shows up at the local level with the
dispersal variable, however, which does not suffer from this problem: loyalty to the
Hum Nat (2013) 24:59–75 71
Author's personal copy

local community was stronger in societies where residential moves between commu-
nities were frequent.
Taken together, the results did not provide the strong and consistent corre-
lations found in recent cross-national pathogen studies. The exception is the
patterning between pathogens and our proxy for collectivist values (socialization
for obedience and self-reliance), which was robust. This result is consistent
with the pathogen theory of group bias, but a more direct interpretation may be
that adults demand obedience and do not encourage self-reliance so as to
protect children from sources of disease and disease vectors where these
dangers are high.
Our negative results could be due to limitations in our data and/or real differences
in the populations being studied. We discuss these two issues next.
Our ethnographic database has problems of heterogeneity, small sample size,
and poor resolution. The ethnographic world represented in the SCCS is highly
heterogeneous (by design); the societies are more diverse than societies in
cross-national comparisons, and they differ in many ways that are likely to
have independent effects on the dependent variables being measured. We also
had small samples for many of our variables, which makes it difficult to control
for moderators. We had more power with the socialization data than with the
mobility or group bias data because the sample was about twice the size, and
the resolution of the scale was finer. This may be why we had the most
convincing patterning with the child socialization variables.
Our pathogen data also have weaknesses. We used historical data because global
infectious disease prevalence has changed markedly owing to conquest, inter-regional
travel, environmental changes, and advances in public health. Historical sources,
however, are not only coarse-grained and less accurate than modern disease preva-
lence data, they are biased by mode of transmission toward diseases spread by
vectors, as noted above. It is possible that responses of the behavioral immune system
are symptom-specific. If so, we might expect social avoidance responses to be
triggeredmorereadilybydiseasestransmitted through bodily effluvia, such as
diarrhea (fecal-oral transmission) and coughs and sneezes (droplet transmission),
than by diseases spread through insects and other vectors (although some vector-
borne diseases cause disfiguring lesions, which can lead to ostracism). Historical
datasets might then provide a weaker test of the parasite-stress model of human
sociality than modern contemporaneous data. Thornhill et al. (2010) have shown that
their theory holds more strongly for non-zoonotic diseases (those that infect only
humans, whether transmitted through vectors or not). It would be interesting to know
whether stratification by mode of transmission also shows patterning in the behav-
ioral response.
Little is known about the processes that would lead people to develop adaptive
aversions to outsiders in the presence of disease. Schaller and Murray (2010) suggest
that while many practices that reduce disease transmission are likely to have become
prevalent through cultural evolution, “aversive emotions and cognition”have also
evolved through natural selection under the pressure of infectious disease. Cross-
cultural differences in these emotional and cognitive responses could reflect either
differences in gene frequencies or the differential expression of a shared genotype to
different conditions (i.e., a similar norm of reaction to variation in disease stimuli). In
72 Hum Nat (2013) 24:59–75
Author's personal copy

either case, however, it is likely that culturally transmitted theories of illness will
amplify or reduce these evolved propensities.
The difference between our results and those of other studies could therefore
reflect differences in the culturally transmitted theories of illness held by
traditional and modern educated, industrialized populations. One does not need
a germ theory of disease to recognize that that one can get sick by being near
sick people (contagion), and a theory of contagion would provide conscious
reinforcement for emotional avoidance reactions to illness. It might seem as
though this awareness would be universal, but it is not: contagion as a source
of illness is not widely accepted by most of the societies in the SCCS. In an
analysis of cultural theories of illness among the SCCS societies, Murdock et
al. (1978) show that people in these societies do not think they get sick by
catching diseases from other people. The most important causes of illness were
thought to be spirit aggression and sorcery, and most societies were found to
have no concept of contagion as a cause of disease, even using a definition of
contagion that incorporates the supernatural (“coming into contact with some
purportedly polluting objects, substance, or person”). Among the minority
(46/128) of societies that do recognize contagion, all but one thought it to be
a minor or unimportant contributing factor to the primary causes of illness,
which were, as with the full sample, supernatural agents. The same held true
for societies with theories of infection: only a quarter (of 122) had such
theories, and all but one of them considered it minor or relatively unimportant.
We should therefore consider the possibility that the responses predicted by the
pathogen theory of group bias may be expressed more strongly in the contem-
porary world, where an understanding of disease transmission is widespread and
taught to children from an early age. Modern, educated populations (the source
of most social science data) are psychological outliers in many respects
(Henrich et al. 2010); this could be another.
Acknowledgments Our greatest debt is to Damian Murray, for training us in the use of the pathogen
sources he used for his cross-national historical pathogen codes. We thank him and Mark Schaller for their
encouragement with the project and help with the literature. We are grateful to Bobbi Low for both her own
insightful work and her consistent encouragement of our own efforts. We also thank Ryan Bohlander for his
exploratory data analysis and persistence in the face of negative results, and Carol Ember and our
anonymous reviewers for very helpful suggestions.
References
Barry, H., Josephson, L., Lauer, E., & Marshall, C. (1976). Traits inculcated in childhood: cross-cultural
codes 5. Ethnology, 15(1), 83–106.
Barth, F. (1969). Ethnic groups and boundaries: The social organization of culture difference. Boston:
Waveland Press.
Binford, L. (2001). Constructing frames of reference: An analytical method for archaeological theory
building using ethnographic and environmental data sets. Berkeley: University of California Press.
Brewer, M. (1999). The psychology of prejudice: ingroup love and outgroup hate? Journal of Social Issues,
55(3), 429–444.
Cashdan, E. (2001). Ethnocentrism and xenophobia: a cross-cultural study. Current Anthropology, 42(5),
760–765.
Hum Nat (2013) 24:59–75 73
Author's personal copy

Cashdan, E. (2012). In-group loyalty or out-group avoidance? Isolating the links between pathogens and in-
group assortative sociality. The Behavioral and Brain Sciences, 35(2), 82.
Faulkner, J., Schaller, M., Park, J., & Duncan, L. (2004). Evolved disease-avoidance mechanisms and
contemporary xenophobic attitudes. Group Processes and Intergroup Relations, 7(4), 333.
Faust, E., & Russell, P. (1964). Craig and Faust’s clinical parasitology. Philadelphia: Lea and Febiger.
Fincher, C., & Thornhill, R. (2008a). Assortative sociality, limited dispersal, infectious disease and the
genesis of the global pattern of religion diversity. Proceedings of the Royal Society B: Biological
Sciences, 275(1651), 2587–2594.
Fincher, C., & Thornhill, R. (2008b). A parasite-driven wedge: infectious diseases may explain language
and other biodiversity. Oikos, 117(9), 1289–1297.
Fincher, C., & Thornhill, R. (2012). Parasite-stress promotes in-group assortative sociality: the cases of
strong family ties and heightened religiosity. The Behavioral and Brain Sciences, 35(2), 39–59.
doi:10.1017/S0140525X11001774.
Fincher, C., Thornhill, R., Murray, D., & Schaller, M. (2008). Pathogen prevalence predicts human cross-
cultural variability in individualism/collectivism. Proceedings of the Royal Society B: Biological
Sciences, 275(1640), 1279–1285.
Gelfand, M., Raver, J., Nishii, L., Leslie, L., Lun, J., Lim, B., Duan, L., Almaliach, A., Ang, S., Arnadottir,
J., et al. (2011). Differences between tight and loose cultures: a 33-nation study. Science, 332(6033),
1100.
Henrich, J., Heine, S., & Norenzayan, A. (2010). The weirdest people in the world. The Behavioral and
Brain Sciences, 33(2–3), 61–83.
Low, B. (1989). Cross-cultural patterns in the training of children: an evolutionary perspective. Journal of
Comparative Psychology, 103(4), 311–319.
Low, B. (1990). Marriage systems and pathogen stress in human societies. American Zoologist, 30(2), 325–
340.
Low, B. S. (1994). Pathogen intensity cross-culturally: SCCS. World Cultures,8(2).
Moody, J. (2001). Race, school integration, and friendship segregation in America. The American Journal
of Sociology, 107(3), 679–716.
Murdock, G. P., & Provost, C. (1971). Measurement of cultural complexity. Ethnology, 12, 379–392.
Murdock, G., & White, D. (1969). Standard cross-cultural sample. Ethnology,8(4), 329–369. Updated at
http://128.195.148.16/~drwhite/pub/SCCS1969.pdf.
Murdock, G. P., & Wilson, S. F. (1972). Settlement patterns and community organization: cross cultural
codes 3. Ethnology, 11, 254–295.
Murdock, G., Wilson, S., & Frederick, V. (1978). World distribution of theories of illness. Ethnology, 17(4),
449–470.
Murray, D., & Schaller, M. (2010). Historical prevalence of infectious diseases within 230 geopolitical
regions: a tool for investigating origins of culture. Journal of Cross-Cultural Psychology, 41(1), 99–
108.
Navarrete, C., & Fessler, D. (2006). Disease avoidance and ethnocentrism: the effects of disease vulner-
ability and disgust sensitivity on intergroup attitudes. Evolution and Human Behavior, 27(4), 270–282.
Putnam, R. (2007). E pluribus unum: civic engagement in a diverse and changing society. Scandinavian
Political Studies, 30(2), 137–174.
Rodenwaldt, E., & Bader, R. E. (1961). World-atlas of epidemic diseases (1952–1961). Hamburg: Falk-
Verlag.
Ross, M. (1983). Political decision making and conflict: additional cross-cultural codes and scales.
Ethnology, 22, 169–192.
SAS Institute (2012). Base SAS(R) 9.3 procedures guide:Statistical procedures. SAS Institute.
Schaller, M., & Duncan, L. (2007). The behavioral immune system: Its evolution and social psychological
implications. In J. P. Forgas, M. G. Haselton, & W. von Hippell (Eds.), Evolution and the social mind:
Evolutionary psychology and social cognition (pp. 293–307). New York: Psychology Press.
Schaller, M., & Murray, D. (2008). Pathogens, personality, and culture: disease prevalence predicts
worldwide variability in sociosexuality, extraversion, and openness to experience. Journal of Person-
ality and Social Psychology, 95(1), 212.
Schaller, M., & Murray, D. (2010). Infectious diseases and the evolution of cross-cultural differences. In M.
Schaller (Ed.), Evolution, culture, and the human mind (pp. 243–256). New York: Psychology Press.
Shipley, B. (2002). Cause and correlation in Biology:A user’s guide to path analysis,structural equations
and causal inference. Cambridge University Press.
Simmons, J. S., Whayne, T. F., Anderson, G. W., & Horack, H. M. (1944). Global epidemiology: A
geography of disease and sanitation. Philadelphia: J. B. Lippincott.
74 Hum Nat (2013) 24:59–75
Author's personal copy

Skitka, L. (2005). Patriotism or nationalism? Understanding post-September 11, 2001, flag-display behav-
ior. Journal of Applied Social Psychology, 35(10), 1995–2011.
Thornhill, R., Fincher, C., Murray, D., & Schaller, M. (2010). Zoonotic and non-zoonotic diseases in
relation to human personality and societal values: support for the parasite-stress model. Evolutionary
Psychology, 8(2), 151–169.
Whiting, J. W. M. (1989). Climate data from weather stations. World Cultures, 1(1), 179–199.
Elizabeth Cashdan is a professor of anthropology at the University of Utah. Her research has included a
variety of topics in human behavioral ecology and evolutionary psychology. Her current research focus is
cross-cultural cognition, particularly spatial cognition in foraging societies. She has done fieldwork in
Botswana, Tanzania, and Tonga.
Matthew Steele graduated from the University of Utah with a B.S. in anthropology with a focus in
evolutionary psychology. His research interests include the origin of god concepts, nonhuman animal
cognition, and behavioral immune systems in small-scale societies.
Hum Nat (2013) 24:59–75 75
Author's personal copy