Content uploaded by Martin Lang
Author content
All content in this area was uploaded by Martin Lang on Aug 19, 2021
Content may be subject to copyright.
Evolution and Human Behavior xxx (xxxx) xxx
Please cite this article as: Dimitris Xygalatas, Evolution and Human Behavior, https://doi.org/10.1016/j.evolhumbehav.2021.05.006
1090-5138/© 2021 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Social inequality and signaling in a costly ritual
Dimitris Xygalatas
a
,
b
,
*
, Peter Maˇ
no
c
,
d
, Vladimír Bahna
c
, Eva Kundtov´
a Klocov´
a
d
,
Radek Kundt
d
, Martin Lang
d
, John H. Shaver
e
,
f
a
Department of Anthropology, University of Connecticut, Storrs, USA
b
Department of Psychological Sciences, University of Connecticut, Storrs, USA
c
Institute of Ethnology and Social Anthropology, Slovak Academy of Sciences, Bratislava, Slovakia
d
LEVYNA Laboratory for the Experimental Research of Religion, Masaryk University, Brno, Czech Republic
e
Religion Programme, University of Otago, Otago, New Zealand
f
Centre for Research on Evolution, Belief and Behaviour, University of Otago, Otago, New Zealand
ARTICLE INFO
Keywords:
Social status
Costly signaling
Mauritius
Ritual
ABSTRACT
Evolutionary perspectives suggest that participation in collective rituals may serve important communicative
functions by signaling practitioners' commitment to the community and its values. While previous research has
examined the effects of ritual signals at the individual and collective level, there has been limited attention
directed to the impact of socio-environmental factors on the quality of ritual signaling. We examined this impact
in the context of the Thaipusam Kavadi, a collective ritual performed by Tamil Hindus worldwide that involves
body piercings and other costly activities. We show that participants' relative position in the social hierarchy
systematically affects the form of ritual signaling. Specically, we found that low-status participants are more
likely to engage in signaling modalities that require somatic and opportunity costs in the form of body piercings
and cumulative effort, while high-status individuals are more likely to use nancial capital, in the form of more
elaborate material offerings to the deity. Moreover, signaling in each particular modality is stronger among
individuals who participate in more public (but not private) rituals, corresponding to their long-term commit-
ment to the community. In sum, our results demonstrate that social hierarchies exact unequal requirements on
ritual participants, who in turn modify their signaling strategies accordingly.
1. Introduction
In recent decades, evolutionary perspectives have increasingly
sought to explain some often extravagant and apparently wasteful as-
pects of human behavior through the lens of signaling theory. Stemming
from observations of animal behavior (Darwin, 1859; Zahavi, 1975;
Zahavi & Zahavi, 1997), signaling theory argues that many ostensibly
wasteful behaviors function to reliably communicate information about
the underlying qualities of the actor that might be otherwise hard to
discern. This framework has been productively applied to various do-
mains of human behavior, including the practice of religious rituals
(Boone, 2000; Cronk, 1994a, 1994b, 2005; Iannaccone, 1994; Irons,
2001; Power, 2017a; Shaver & Sosis, 2018; Smith & Bliege Bird, 2000;
Sosis, 2000; Xygalatas, 2012).
Specically, the costs of performing certain religious rituals may
provide a reliable signal of loyalty to the religious group and commit-
ment to the community's beliefs, norms, and values (Sosis & Alcorta,
2003). Thus, the effective communication of a person's commitment can
enhance the practitioners' reputation (Power, 2017b); deter free-riders
(Bulbulia, 2004; Iannaccone, 1994); facilitate cooperation (Sosis &
Rufe, 2003); and increase the credibility of the beliefs associated with
these rituals (Henrich, 2009). In recent years, a wealth of empirical work
has examined the reception of ritual signals by the community and the
social benets of participation in costly rituals (Power, 2017a, 2017b;
Power, 2018 Soler, 2012; Sosis & Bressler, 2003; Sosis & Rufe, 2003;
Xygalatas et al., 2013; Xygalatas et al., 2017). Nevertheless, less atten-
tion has been paid to contextual factors that may impact the production
of ritual signals.
The theory motivating this body of work is known as costly signaling
theory; however, critics have noted that this terminology has led to
confusion for various reasons (Maynard Smith & Harper, 2003).
Notably, the costliness of a behavior does not necessarily mean that it
functions as a signal. Rather, what is critical to ensure signal reliability is
that signal costs and benets of religious behavior are differentially
* Corresponding author at: Department of Anthropology, University of Connecticut, Storrs, USA.
E-mail address: xygalatas@uconn.edu (D. Xygalatas).
Contents lists available at ScienceDirect
Evolution and Human Behavior
journal homepage: www.elsevier.com/locate/ens
https://doi.org/10.1016/j.evolhumbehav.2021.05.006
Received 4 March 2019; Received in revised form 30 April 2021; Accepted 20 May 2021
Evolution and Human Behavior xxx (xxxx) xxx
2
outlaid between low- and high-quality signalers, or that under condi-
tions in which high- and low-quality signalers pay similar costs, high-
quality signalers benet more than low-quality ones (Grafen, 1990;
Maynard Smith & Harper, 2003). While signaling theory has generated a
productive line of research in behavioral ecology and evolutionary an-
thropology, there has been little empirical research on how costs may be
differentially distributed across actors of varying quality (Grose, 2011).
Human signals are embedded in complex socio-cultural environ-
ments, where the costs and benets of signal production may vary be-
tween different senders, receivers, and situations, and therefore can be
difcult to quantify (Barker, Power, Heap, Puurtinen, & Sosis, 2019).
For example, as different individuals are endowed with unequal
amounts of material, social, and embodied capital, such inequalities can
provide divergent affordances, opportunities, and cost-benet ratios
across different signaling modalities. Socio-ecological factors such as
status and wealth inequalities should therefore inuence a sender's
perceived quality and impact signal cost (Buser, 2015; Iannaccone,
1998; Shaver, 2015; Shaver & Sosis, 2014). To ensure the reliability of a
religious signal, then, this would either mean that those of relatively
high socio-economic status (SES) would need to pay lower nancial
costs to demonstrate the same level of commitment than those of low
SES, or that when all signalers may assume similar costs, those of high
status may benet more (Shaver, 2015; Shaver & Sosis, 2014).
In this study, we focus on the intensity of signaling by examining how
socio-economic status affects the cost of religious signals. Specically, in
the context of a collective ritual that requires multiple types of in-
vestments, we investigate whether participants of different socio-
economic status would assume differential costs. Moreover, as ritual
costs can be paid in various currencies (Barker et al., 2019), we quantify
the differential investments of somatic and extra-somatic resources into
signal production across socio-economic strata.
1.1. Ethnographic setting
Our study was conducted in Mauritius, a tropical island in the Indian
Ocean where our team has been engaged in ethnographic eld work
since 2009. Mauritius is one of the most diverse societies in the world. As
it was one of the world's last sovereign countries to be settled by humans
(being continuously inhabited for only three centuries), there is no
indigenous population. The local inhabitants derive their ancestry from
numerous ethnic groups who arrived as a result of colonial policies that
introduced successive waves of incoming populations—rst slaves and
later indentured laborers brought to work on sugar plantations. Those
multi-ethnic encounters resulted in a mosaic of religious beliefs and
practices. The largest religious communities on the island include
Hindus and Muslims (mostly of Indian origin), as well asChristians
(predominantly of African or European descent). Smaller groups include
Buddhists and practitioners of various forms of Chinese folk religion
(typically of Chinese origin), and numerous other faiths (Statistics
Mauritius, 2012). All of those groups are internally heterogeneous and
include adherents of numerous smaller denominations and sub-groups
(Xygalatas, 2013).
Mauritian Hindus, who make up almost half of the population, are
divided into several ethno-religious entities, the largest of which in-
cludes Hindi-speaking Hindus of North Indian origin. Other groups
include Telugus (from Southern India), Marathis (from the Indian state
of Maharashtra), and Tamils (from the state of Tamil Nadu). The latter
are the second-largest Hindu community, comprising about 12% of all
Mauritian Hindus. Tamil temples are found in every major town in
Mauritius, characterized by their colorful frescoes and elaborate
Dravidian architecture. These temples host various physically
demanding rituals that involve painful activities such as re-walking,
sword-climbing, and skin-piercing (Fischer et al., 2014). The most
intense of those rituals is the Thaipusam Kavadi.
Every year, on the tenth month of the Tamil calendar, Tamil Hindus
celebrate the Thaipusam festival in honor of Lord Murugan, the Hindu
deity also known as Kartikeya. This festival is connected to ancient
Puranic legends narrating Murugan's epic battles with demonic forces
(asuras). It commemorates the occasion on which his mother, Parvati,
gave him the spear (vel) that he used to subdue the evil demon Sur-
apadman (Xygalatas, 2021). This is why Murugan is typically depicted
holding a spear, as well as one of the mythical etiologies for why the
Kavadi ritual involves bodily piercing.
During the festival, devotees undergo various physical and mental
austerities, such as following a vegan diet, abstaining from sexual ac-
tivities and alcohol, bathing in cold water, or sleeping on the oor. The
festival culminates on the full moon of the month of Thai, when par-
ticipants carry heavy structures called kavadi (burdens) to the temple of
Murugan on a procession that lasts several hours. This is why the ritual
itself is also called Kavadi.
1
Before embarking on this procession, pil-
grims have their body pierced by numerous needles, skewers, and other
metallic objects. The Thaipusam Kavadi is one of the world's most
widely practiced collective rituals, performed by millions of Tamils and
other Hindus in India and Sri Lanka, and by Diasporic Hindu commu-
nities around the globe (Belle, 2017).
The data for the current project were collected in Quatre Bornes, a
town of about 70,000 inhabitants located in the district of Plaines Wil-
hems, which is home to the largest Tamil population in Mauritius and
home to several Tamil temples. The most important among those tem-
ples is the Sri Siva Subramanya Kovil, more commonly known as Kovil
Montagne (Mountain Temple), which was the rst temple in Mauritius
to host the Thaipusam Kavadi ritual. Built at the foot of the Corps de
Garde mountain in 1907, it gradually grew from an initial small shrine to
a large concrete structure decorated with hundreds of colorful sculptures
and frescoes and is now a major pilgrimage site in Mauritius, especially
during the days of the Kavadi ritual.
The festival is the largest and most important religious event of the
Tamil community, and the day of Thaipusam is a public holiday in
Mauritius. On that day, the streets are lled with massive crowds
forming numerous processions, each leading to a different temple. The
Kovil Montagne is home to one of the largest processions on the island.
The festivities are attended widely, not only by Tamils but by all local
Hindu groups, and even some Christians and members of other religious
communities (Xygalatas et al., 2016).
Over the days leading up to Thaipusam, devotees prepare by fasting
and praying. Each night, they stay up late preparing their kavadis, which
are built on a bamboo, wooden, or metal frame and adorned with
owers and peacock feathers. On the day of Thaipusam, participants
gather at a riverbank at the edge of the town to bathe and perform
cleansing rites. The kavadis are displayed next to their owners, who have
their tongues pierced by silver needles with the help of a priest. While
most women have a single piercing, some of the men have hundreds of
needles impaled throughout their torso and limbs, as well as hooks from
which they hang lime fruit or bells. Additionally, many of the men have
their cheeks pierced by aluminum skewers, steel rods, and other long
spears. Some also attach hooks through the skin of their back, by which
they drag large chariots through chains. Once these piercings are in
place, they lift their kavadis and balance them on their shoulders before
embarking on a pilgrimage to the temple of Murugan.
The pilgrimage follows a 6 km-long route that takes many hours to
cover, as the procession pauses at every crossroad. At each stop, priests
perform cleansing rites and musicians play drums and the Nadaswaram
(a traditional Tamil wind instrument). As the rhythm picks up, many
devotees appear to fall into trance while dancing in a swirling fashion
with their kavadis still perched on their shoulders. Throughout this
1
In this article, we use the term capitalized (Kavadi) to refer to the ritual, and
in lower-case (kavadi) to mean the structure carried by participants in that
ritual. Note that there is no standard transliteration scheme for the romaniza-
tion of Tamil. As a result, there are numerous variations in the English spelling
of both Thaipusam and kavadi, and our choice is simply a matter of preference.
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
3
procession, pilgrims walk barefooted on the scorching asphalt or wear
sandals made of upwards facing nails. During this time, they never put
their burden down and are not allowed any food or water. When they
nally reach the temple of Murugan, they lay down their kavadi and
have their piercings removed, but only after tackling one last hardship:
as the Kovil Montagne is located on a mountainside, devotees must
climb 242 steps to carry their burden to the top.
Our ethnographic interviews revealed a multitude of interpretations
with regards to the performance of this ritual (Xygalatas & Maˇ
no, 2021).
Participation is often related to a vow made to the deity in exchange for
a favor or is offered as a token of gratitude for some good fortune already
received (Ward, 1984). But although specic benets of the ritual such
as health and healing were often mentioned (Xygalatas et al., 2019),
most responses referenced tradition, as reected in statements such as
“we do it because it is our tradition” or “because we are Tamils”; “we do
it for Murugan”, or “because we believe”. Despite the lack of a clear
justication for the ritual, however, the overwhelming majority of our
informants afrmed that the Kavadi was the most important ritual event
in their lives.
Participation in the Kavadi involves high and multifaceted costs that
spread across various domains. Among others, embodied costs include
the severe pain and stress of enduring multiple piercings as well as
substantial health risks such as injury, trauma, and infection. Financial
costs are also signicant, as participants spend a sizable portion of their
income (typically ranging between a few days' and several weeks' salary)
to build their kavadis and lose additional income by taking days off work
during the festival. Finally, opportunity costs include the time spent
participating in the festival, which can accrue substantially over its
duration, thus hindering the ability to invest in other domains. Ac-
cording to our estimates, attending all temple activities during the
festival requires over 30 h (not including time spent commuting), while
participants reported spending an additional 19 h in other preparations
on average. Over time, this can result in an investment of several months
during one's lifespan. Willingness to undertake the pilgrimage can thus
be ostensibly seen as an honest and reliable signal of commitment, as
doing so should be costly enough to deter anyone who is not genuinely
devoted. Besides, in addition to the physical risk involved in the ritual,
participation also carries reputational risks. Once initiated, failure to
complete this demanding ritual can bring shame and suspicion in the
eyes of the community. Thus, should participants signal beyond their
capacity and fail, they would not only miss out on benets proportional
to the invested effort, but may also bear additional costs due to that
failure.
2
The decision to participate is thus not to be taken lightly, and
this risk is an additional deterrent to those not fully committed. In point
of fact, previous studies conducted in Mauritius suggest that participa-
tion in this ritual is a predictor of ingroup prosociality (Xygalatas et al.,
2013; Xygalatas et al., 2017).
Although the content of the signal is similar across different groups,
the particular forms of signaling employed can vary across socio-
ecological parameters. Moreover, the costs of participation are not
equal for all participants. Some can have a single piercing but others
endure hundreds; some carry smaller and simpler kavadis and others
carry much larger and more decorated structures; and while some
perform the ritual once in a lifetime, others may actively participate
every year. Based on the above qualities, the Thaipusam Kavadi pro-
vided an ideal setting for testing our hypotheses.
1.2. Predictions
The Kavadi ritual consists of multiple modalities. Previous work
suggests that different modalities may constitute different currencies
whose exchange rates vary across socio-economic parameters. For
example, the opportunity cost of time increases while nancial costs
decrease as a function of wealth (Iannaccone, 1992, 1998). In this
context too, participants may strive to maximize their signaling poten-
tial by utilizing the most efcient signaling modalities available to them.
Specically, we predicted that (1) participants of low socio-economic
status, who lack the means to make substantial nancial investments,
will engage more heavily in those signaling modalities that entail more
embodied and opportunity costs to enhance their cooperative affor-
dances. Therefore, they will perform the ritual more frequently and will
endure a larger number of piercings.
On the other hand, (2) participants of high socio-economic status,
who have more means to make alternative forms of investment (which
are less painful and carry lower risks compared to body-piercing), will
instead engage more heavily in those signaling modalities that require
spending extra-somatic forms of capital. Therefore, they will carry larger
and more elaborate kavadis during the procession, which require better
and more expensive materials to build and decorate.
Moreover, we examine how the signals transmitted during the
Thaipusam Kavadi ritual relate to regular commitment displays
throughout the year. If Kavadi participation is a signal of commitment to
the religious community, this signal should be stronger for more
committed members, that is, those who pay costs to engage in public
rituals also throughout the year. We therefore predicted that (3) in-
dividuals who more frequently engage in public (compared to private)
rituals will produce costlier signals across the modalities more readily
available to them according to their socio-economic status. That is, we
expected that higher public ritual participation throughout the year
would negatively interact with SES in predicting the number of piercings
and Kavadi frequency, but that this interaction would be positive for
kavadi size.
2. Materials and methods
2.1. Participants and procedure
We recruited 80 participants (52 males, 28 females) ranging from 18
to 60 years of age (mean =33.04, SD =11.18), who took part in the
ritual and carried a kavadi. Recruitment took place on the day of
Thaipusam. On that day, the procession follows a pre-specied route,
beginning at the Northern entrance of Quatre Bornes and ending at the
temple of Murugan. We recruited participants at both of those locations
to avoid interrupting the procession. Local research assistants
approached participants who had kavadis and invited them to partici-
pate in the study as they waited for the procession to begin or as they
were exiting the temple. Participants signed informed consent forms and
written permission was obtained by the temple committee.
2.2. Ritual costs
On the day of the ritual, we recorded the number of piercings for
each participant, including needles, hooks, and skewers impaled
through their body. In addition, we used measuring tapes to record the
height, width, and depth of each participant's kavadi. The three di-
mensions were then multiplied to calculate the volumetric size of the
structure in terms of cubic capacity. Moreover, participants reported
how many times they had performed the ritual during their lifetime.
Over the following two weeks, our team visited each participant at their
home and administered a demographic survey. This was done to mini-
mize intrusion on the day of the ritual and to avoid creating a bottleneck
at the point of data collection.
2.3. Demographic data
Our surveys included socio-demographic information (age, sex,
occupation, marital status) and data on religiosity. Specically, partic-
ipants were asked to rate their level of religiosity on a 10-point scale,
2
We thank one of the anonymous reviewers who pointed this out during the
revision process.
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
4
with 1 being “I am the least religious person I know” and 10 being “I am
the most religious person I know”. Moreover, they reported on the fre-
quency of public ritual (attending events at the temple) and private
ritual practice (performing rituals at their own home). Those data were
then recoded to reect daily frequency (see SM).
To assess socio-economic status (SES), we used a composite index
comprising of six demographic variables related to material and cultural
capital (Shavers, 2007; Vyas & Kumaranayake, 2006): occupational
prestige, education, English language prociency, income, and real es-
tate and car ownership. For occupational prestige, we compiled an
exhaustive list of participants' occupations and had them rated by 20
independent local raters unrelated to the current sample on a 5-point
scale ranging from “not at all prestigious” to “very prestigious”. Those
ratings were then averaged to provide a measure of occupational pres-
tige (ICC: 2, k =0.951).
Educational achievement was measured by two variables: years
spent in formal education, and level of English language competency as
reported by participants. English is the language of business and
administration in Mauritius, and because it is not identied with any
particular ethnic group it is a good indicator of education as well as
status (Sauzier-Uchida, 2009), having the highest level of prestige of all
the languages spoken on the island and acting as a gateway to economic
opportunities and social mobility (Sauzier-Uchida, 2009).
Because asking direct questions about one's income is often
perceived as inappropriate in Mauritius (and may result in inaccurate
reporting), we used occupation as a proxy for income. The same 20 in-
dependent local raters were asked to estimate the typical monthly in-
come for each profession, and those estimates were averaged to obtain a
measure of monthly income (ICC: 2, k =0.757). As expected, the in-
dicators of occupational prestige and income were closely related, and a
gamma distribution appeared to best describe this relationship (β =
1.16, [0.85,1.46]; see Fig. S2 for details).
Finally, we obtained information on material property by using two
variables: ownership of real estate (immovable property), measured as
the number of houses and plots of land owned by the nuclear family; and
ownership of vehicles, measured as the number of cars owned by the
nuclear family. Car ownership in Mauritius is not only an indicator of
family afuence but also a potent facilitator of employment opportu-
nities, connectivity, and social mobility (Pendall et al., 2014).
Since 13 participants missed records on at least one of these SES
indicators (2% of data on predictor variables missing), we imputed the
missing data using predictive mean matching from the mice package
(Van Buuren & Groothuis-Oudshoorn, 2011). The six SES indicators
with imputed data were then subjected to an exploratory factor analysis
with an expected one-factor solution. The factor analysis revealed
loadings ranging from 0.55 to 0.84, with a Cronbach's alpha of 0.79,
McDonald's total omega of 0.89, and hierarchical omega of 0.52. The
SES indicators were then factor-scored to produce a composite measure
of SES that ranged from approx. −2 to 3. See Supplementary Material for
more details on the demographic data, imputations, factor analysis and
factor scoring.
2.4. Analysis
All analyses were conducted in R, version 3.6.3 (R Core Team, 2019)
using functions glm and glm.nb. Plots were created using the package
ggplot2 (Wickham, 2016). Anonymized data and code are available as
part of the Supplemental Materials.
To evaluate our hypotheses, we built a series of Generalized linear
models (GLMs) for each of our measures of ritual intensity: number of
piercings; Kavadi frequency; and kavadi size. Since the number of
piercings and the number of previous Kavadi performances indicated a
distribution typical for count data, we used negative binomial regression
with a log link to model the effects of SES on these two variables. Kavadi
size, on the other hand, was best characterized by a non-negative
continuous distribution, and a gamma distribution with a log link
tted well to those data.
For all dependent variables, we rst included only the SES variable to
estimate its unadjusted effects, and in subsequent steps added the de-
mographic and religiosity variables. We also considered including
squared age effects, but since age had only very weak effects (see the
results section), this term was not included in the models. In the nal
step, we added an SES*Public ritual frequency interaction term to esti-
mate how the differential modes of ritual intensity correspond to regular
commitment signals. However, since the public ritual frequency data
included three outliers (higher than 3 times the interquartile range plus
the third quartile value) that pushed the model estimates to unrealistic
predictions, we also report models excluding those three outliers. The
results of models with and without outliers lead to the same inference on
the relationship between ritual intensity, SES, and public ritual fre-
quency, but the latter models yield more realistic estimates. Finally,
since we expected that the effects would be driven mostly by males, we
built the same models only for males.
In the main text, we report only basic models including SES, models
with demographics and religiosity, and models with the SES*Public
ritual interaction that exclude the tree outliers. All modeling steps are
reported in the Supplementary Material. Likewise, we report untrans-
formed raw estimates with 95% CI from the negative binomial and
gamma models in the text and tables, but predicted values are plotted in
the gures located in the results section. We also provide bootstrapped
95% CI for the main results (effects of SES and SES*Public ritual inter-
action). Condence intervals based on parametric bootstrap are dis-
played in Fig. S8.
3. Results
3.1. Sample description
Before examining the relationship between the main predictor vari-
ables (SES, public ritual frequency) and outcome variables (number of
piercings, Kavadi frequency, kavadi size), we investigated the relation-
ship between demographic and religious control variables that dened
our sample. The relationship between SES and demographic and reli-
giosity variables was rather weak (Pearson's r with age: −0.22; religi-
osity: 0.14; private ritual performance: −0.01; public ritual
performance: −0.06), suggesting that religiosity and ritual practice in
general were relatively stable across different socio-economic strata (see
Fig. S1). Interestingly, the inter-correlations between religiosity, private
ritual practice, and public ritual performance were also weak, which is
likely due to the modes of these variables being close to ceiling/oor
levels (see density plots in Fig. S1). This is especially true for public
ritual performance, which has most values within the interval between
0 and 0.5, yet there are three outliers with values of 1 and 2 (performing
public rituals daily and twice a day, respectively). Since these outliers all
lay within the lower SES strata (see Fig. S4), they substantially bias the
estimates of the relationship between SES and public ritual performance.
For this reason, we present models with and without those outliers (see
Analysis).
We further explored the differential associations of our variables
between males and females, as this comparison was paramount to our
decision to focus specic analyses on the male sample. There were no
substantial differences between females and males in the inter-
correlations of our predictor variables except for age, with females
having a more negative relationship between age and SES (Pearson's r =
−0.54) than males (Pearson's r = − 0.08). This negative relationship is
driven predominantly by the lower levels of education and English
prociency among elder women in our sample (Pearson's r of age and
years of education = − 0.65, English prociency = − 0.43). Moreover,
men are more likely to join the workforce, as some women still
(although decreasingly so) remain at home as housewives or work in
low-status jobs, with men holding occupations that pay over twice as
much (29,606 MUR, SD =28,036) as those for women (14,739 MUR, SD
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
5
=18,052). Also see Fig. S3 for histograms of income and occupational
prestige broken down by sex. There were further differences between
males and females in religiosity and ritual performance; however, these
differences were likely caused by the low sample size of female partic-
ipants and were imprecisely estimated, as revealed by the scatterplots in
Fig. S1. Also see Fig. S4 for histograms of religiosity variables broken
down by sex.
In the next step, we inspected the differences in our outcome vari-
ables between males and females. We found major differences between
sexes in terms of their participation in the Kavadi, as men's engagement
in this ritual was much more intense relative to women (see Fig. 1). On
average, the men in our sample endured 78.69 piercings (SD =129.48)
compared to only 1.11 (SD =1.03) for women. Similarly, the average
size of kavadi carried was 1.65m
3
for men (SD =1.44) and 0.69m
3
(SD
=0.64) for women. These differences were in line with our ethnographic
observations that the extreme forms of the ritual are exclusive to men,
and therefore males present a better population for testing our hy-
potheses. In the next section, we test the robustness of these observations
with generalized linear models.
3.2. Kavadi frequency
We predicted that SES would be negatively associated with Kavadi
frequency, as lower-SES participants would use this ritual platform more
often to signal commitment. In our sample, eight subjects took part in
the Kavadi ritual for the rst time while the highest number of previous
performances was 36. Using a negative binomial model, we found that
SES negatively predicted kavadi participation frequency (β = − 0.24,
[−0.46, −0.01]; Table 1, Model 1). This basic model predicted an
average of four previous participations for those with the highest SES,
compared to ten previous participations for those with the lowest SES.
Adjusting the model for demographic and religiosity variables, the SES
estimate remained practically the same (β = − 0.23, [−0.44, −0.01]).
The interaction between SES and public ritual was positive, although
this coefcient was imprecisely estimated (β =0.80, [−1.34, 2.94]). See
Table 1, Model 3 and Fig. 2A for a visual illustration of the interaction
between SES and public ritual. Surprisingly, age did not explain a sub-
stantial amount of variance in previous kavadi participation (β =0.01,
[−0.01, 0.03]). On the other hand, religiosity was a stable positive
predictor of Kavadi frequency across all model specications (β =0.11,
[0.02, 0.20]). As expected, males had participated more frequently than
females in previous Kavadi rituals (Table 1, Model 2).
Repeating these analyses for the male sample, we found that SES was
a stronger predictor of Kavadi frequency than in the full sample (β =
−0.35, [−0.61, −0.08]). This basic model that included only SES pre-
dicted that participants with the lowest SES took part in approximately
15 Kavadi rituals compared to four previous participations for those
with the highest SES. In the male sample, age was a positive predictor of
the number of previous Kavadi performances (β =0.04, [0.01, 0.06]).
Finally, as in the full sample, the interaction between public ritual and
SES was positive but imprecisely estimated (β =1.21, [−1.03, 3.45]).
3.3. Number of piercings
We predicted that SES would be negatively associated with the
number of piercings, as lower-SES participants would use their somatic
capital to signal commitment. The lowest observed number of piercings
was 0, the median number was 110, and the highest number was 600
piercings, with a distribution typical of count data (see Fig. 1). Using a
negative binomial model, we found that SES negatively predicted the
number of piercings (β = − 0.33, [−0.75, 0.08]), although the 95% CI
crossed zero (Table 2, Model 1). This basic model predicted 23 piercings
for participants with the highest SES compared to around 86 piercings
for participants with the lowest SES. Adjusting the model for de-
mographic and religiosity variables did not substantially change the
estimated SES effect (β = − 0.26, [−0.61, 0.09]). Further investigating
the interaction between SES and public ritual performance yielded a
strong negative coefcient (β = − 3.50, [−6.85, −0.14]), suggesting that
for lower-SES participants frequent public ritual performance predicted
an increasing number of piercings. See Table 2, Model 3 and Fig. 2C for a
visual illustration of the interaction between SES and public ritual.
Covariates included in these models did not reveal stable effects except
for sex. Holding all other predictors at their means, males were predicted
to have on average 90 piercings while females only two (β =4.09,
[3.32,4.86]).
Since sex appeared to be the strongest predictor of the number of
piercings, we also performed the same analyses on the male sample only.
The results were indeed stronger, with the basic model that included
only SES predicting that participants with the lowest SES would endure
approximately 151 piercings whereas participants with the highest SES
around 27 piercings. Public ritual again appeared to be an important
moderator of this effect, with a similar negative coefcient size as in the
full sample. To compare the results of the full sample and male sample
only, see Table 2 and Fig. 2D.
Fig. 1. Ritual intensity variables broken down by sex. The plots display scatterplots of raw data overlaid with means and standard errors.
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
6
3.4. Kavadi size
For kavadi size, we predicted opposite effects compared to the
number of piercings, namely that this signaling modality would be uti-
lized primarily by participants of higher SES. In accordance with this
expectation, the results of the gamma model revealed that SES positively
predicted kavadi size (β =0.24, [0.04, 0.44]), with the lowest-SES
participants carrying kavadis around 0.8 m
3
in size, compared to
about 2.2 m
3
for the highest-SES participants (Table 3, Model 1). This
effect, however, was substantially weakened after including the de-
mographic and religiosity variables (β =0.12, [−0.06, 0.31]). An
investigation of the interaction between SES and public ritual perfor-
mance again revealed a strong relationship (β =2.13, [0.40, 3.86]),
suggesting that more frequent participation in public rituals was linked
with larger kavadi size among higher-SES participants (although para-
metric bootstrap of 95% CI indicated high variability of this effect,
suggesting a larger sample would be needed; see Fig. S8). Age was
negatively associated with Kavadi size across the model specications
(β = − 0.03, [−0.04, −0.01]) while males generally carried larger
kavadis (β =0.94, [0.58, 1.29]). See Table 3, Model 3 and Fig. 3E.
Analyzing only the male sample again revealed positive effects of
SES. While the effects of SES on kavadi size were unstable in the full
sample, they were stable in the male-only sample, including the SES*-
Public ritual performance interaction. However, the differences between
the two samples were not as pronounced as in the case of the number of
piercings, suggesting that although men carry larger kavadis, women
too may bear sizeable structures (see Table 3 and Fig. 3E and F).
4. Discussion
Our results show that socio-economic factors play a crucial role in
determining the form and intensity of signaling across different mo-
dalities in the context of the Thaipusam Kavadi. Our measure of religi-
osity predicted the frequency of participation, suggesting that this
measure may be taken as a more general long-term assessment of reli-
gious commitment. However, religiosity did not predict the intensity of
performance, and neither did private ritual practice. We do not know
whether this is the case in general among Mauritian Tamils or specic to
our sample, possibly because participants in the Kavadi are a self-
selected group of highly religious individuals. In either case, this sug-
gests that the observed variance in the use of signaling modalities is, as
we predicted, predominantly driven by situational factors rather than
personality traits related to spirituality or ritualization.
More specically, we found that individuals of lower social stand-
ing—especially males—assume greater somatic and opportunity costs
during the Thaipusam festival by participating more frequently and
enduring more piercings. This suggests that low-status signalers pay
higher costs even when they are equally committed. In contrast, those of
higher socio-economic status follow a different signaling strategy by
building larger kavadis. This requires greater nancial costs. Note,
however, that while these costs may be higher in absolute terms, mon-
etary costs are relatively lower for wealthier individuals (Iannaccone,
1998). Our interviews showed that most people spend between 3000
and 5000 rupees in preparing their kavadi, but some of the more elab-
orate structures may cost up to 40,000 rupees. While for high earners
such expensive kavadis may be a nancial inconvenience, they are an
impossibility for low-income individuals, for whom even the cheaper
kavadis impose a substantial burden.
Rituals like the Thaipusam Kavadi provide arenas for various modes
of signaling. Some of these types of signals are clearer than others. For
example, the body piercings allow for an almost standardized compar-
ison between participants who opt to endure more or fewer of them,
while other activities such as carrying the kavadi leave more space for
ambiguity. This is because one can carry a larger structure by expending
more somatic capital (physical effort) but also by expending more
nancial capital, i.e. paying more money to buy better and more light-
weight materials, such as bamboo or aluminum rather than wood or
iron. As the frame of the structure is covered by owers, feathers, and
other decorative items, this is not readily visible to the observer. Indeed,
we found that both strategies are employed in the Kavadi ritual. It seems
that participants, especially men, invest more nancial resources when
they can afford it, but pay higher physical costs when they cannot, e.g.
by putting more needles in their bodies.
One interpretation for this pattern may be that people attempt to
exaggerate the signal whenever they have the opportunity. In this
respect, the size of a kavadi is a more ambiguous signal than the number
of piercings, as it can be more easily manipulated. However, an alter-
native possibility is that carrying a kavadi signals both physical and
nancial qualities, as it requires both caloric and monetary investments.
From this perspective, then, high-status individuals might be better able
to harness the pluripotency of the ritual signal by advertising both their
devotion and their wealth. On the other hand, low-status participants
are more likely to harness this pluripotency by signaling both devotion
and physical tness (Power, 2017a). To differentiate between these
types of signals, future studies will need to gather ne-grained data on
the audience's evaluation of these markers.
Overall, the multimodality of this ritual may help make commitment
signaling more unambiguous by supplying more reliable information
(Partan & Marler, 1999; Rowe, 1999). The Thaipusam Kavadi is a noisy
event—not just literally, but also in terms of its communicative poten-
tial. As numerous signals are being transmitted simultaneously and the
receivers' attention is limited, individual signalers and their signals can
easily be conated, misread, or distorted. To counteract such noise,
communication systems (human and other) may employ a variety of
strategies (Guilford & Dawkins, 1991; Wiley, 2015). Such strategies may
involve increasing the signal's volume, for example, by using larger and
more decorated kavadis, bigger skewers, or more painful actions, such as
dragging chariots or walking on shoes made of nails; increasing the
signal's contrast and clarity, for example by introducing idiosyncratic
variations in the kavadis or wearing less clothing to make the piercings
Table 1
Untransformed estimates with 95% CI from negative binomial models (with log link) predicting the number of previous Kavadi rituals.
All participants Males only
(1) (2) (3) (4) (5) (6)
Intercept 1.93 (1.70, 2.15) 0.67 (−0.17, 1.51) 0.67 (−0.20, 1.54) 2.14 (1.88, 2.41) 1.17 (0.12, 2.22) 1.26 (0.13, 2.40)
SES −0.24 (−0.47, −0.01) −0.23 (−0.44, −0.01) −0.35 (−0.64, −0.05) −0.35 (−0.61, −0.08) −0.23 (−0.48, 0.02) −0.42 (−0.78, −0.06)
Age – 0.01 (−0.01, 0.03) 0.01 (−0.01, 0.04) – 0.03 (0.005, 0.06) 0.04 (0.01, 0.06)
Sex – 0.52 (0.08, 0.97) 0.53 (0.08, 0.99) – – –
Marital status – 0.05 (−0.46, 0.55) 0.002 (−0.52, 0.52) – 0.05 (−0.63, 0.72) −0.05 (−0.75, 0.66)
Religiosity – 0.11 (0.02, 0.20) 0.11 (0.02, 0.20) – 0.12 (−0.005, 0.24) 0.1 (−0.02, 0.23)
Private ritual – −0.03 (−0.26, 0.20) −0.06 (−0.30, 0.18) – −0.07 (−0.33, 0.18) −0.1 (−0.36, 0.17)
Public ritual – 0.55 (−0.18, 1.29) 1.38 (−0.68, 3.44) – 0.54 (−0.22, 1.30) 1.58 (−0.60, 3.76)
SES*Public ritual – – 0.8 (−1.34, 2.94) – – 1.21 (−1.03, 3.45)
N 80 80 77 52 52 49
Note. SES is Socio-Economic Status; Age is centered at its mean; Reference category for Sex is female. Reference category for Martial status is unmarried.
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
7
more visible; increasing predictability (a feature characteristic of most
public rituals, which take place in a predetermined time and space, and
in front of audiences who know when, where, and what signals to
expect); or increasing redundancy, for example through multiple
piercings, repeated performances, and, crucially, by using multiple
means of conveying the same information. Thus, while some of the ac-
tions involved in the kavadi may individually be susceptible to decep-
tion, on the whole they become less noisy and more robust signals of
honest commitment. As a result, rituals that involve multicomponent
signals may have a selective advantage because they provide receivers
with more opportunities for evaluating the reliability of signalers (Soler,
Batiste, & Cronk, 2014; Sosis, 2003), and this might be one reason
extreme rituals tend to be widely attended by large crowds of observers
(Xygalatas, 2015).
While constrained by such selective pressures, social environments
may often interact with those pressures in orthogonal wayspresenting
low-status individuals with a conundrum: not only are they in greater
need to gain the benets of being perceived as committed, but they also
likely attract less public attention than high-status individuals. There is
therefore more uncertainty with regards to the reception of their signals,
which are in greater risk of being missed or ignored. In the face of such
uncertainty, low-status individuals may be motivated to pay higher costs
to overcome the noise, even if they do not reap higher benets.
Alternative interpretations may focus on the role of rituals in pre-
serving social power structures (Leach, 1954). From this perspective,
ritual participants of various socio-economic backgrounds participate in
Fig. 2. Transformed estimates from the GLMs on frequency of Kavadi performance, number of piercings, and kavadi size predicted by SES and moderated by public
ritual performance. Note that for these plots we divided the SES variable into terciles whereas the statistical models included SES as a continuous variable. A.-B.
Increase in public ritual performance was associated with an increase in the reported Kavadi ritual frequency for the high SES category. However, note that these
differences were imprecisely estimated and this plot is to some extent biased by SES categorization into the three groups. C.-D. Increase in public ritual performance
was associated with an exponential increase in the number of piercings in the low SES category only. E.-F. Increase in public ritual performance was strongly
associated with kavadi size with increasing SES. The strongest association was in the high SES category but there was no association in the low SES category.
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
8
the festival in strictly dened socially stipulated roles, and those roles
prescribe the forms and intensity of their ritual actions. In other words,
to the extent that devotees gain reputational benets from their
participation in the kavadi, they do this by meeting the specic
thresholds needed for their society to recognize their devotion. Hence,
the variability in ritual intensity is not necessarily evidence of greater
commitment, nor is it necessarily indicative of a signaling competition,
but can also be seen as evidence of the acceptance and submission to
social hierarchies by signaling devotion in socially accepted and stipu-
lated ways.
It remains to be seen what effects the availability of cheaper and
lighter materials like plastic will have on the practice of the kavadi. As
these increasingly accessible materials are making it ever-easier to carry
larger structures, the cost and reliability of the signal is effectively
decreased (Cronk, 1994a). We might therefore expect that new forms of
kavadi will emerge that will allow their bearers to convey more un-
equivocal signals. Indeed, some examples have already appeared in
recent years. For example, one particular type of kavadi is attached to
the torso by multiple metallic spikes rather than supported on the
shoulders. This type of structure makes the material less ambiguous and
has an added degree of difculty and pain. Alternatively, we might see
kavadis become less important and more emphasis placed on the num-
ber of piercings.
The fact that only men engage in the extreme forms of the kavadi
suggests that this ritual may function (at least partly) as a male display of
mating qualities (Darwin & Wallace, 1858; Smith, Bliege Bird, & Bird,
2003). Our current data do not speak to this possibility, which might be
explored in further studies. However, the Thaipusam festival is not by
any means limited to those painful actions. Participation and signaling
in the context of this tradition may come in many forms, only some of
which we quantied in this study. For example, women may often
engage in different signaling strategies than men (Bliege Bird & Smith,
2005; Bulbulia, Shaver, Greaves, Sosis, & Sibley, 2015; Shaver, 2015). In
the context of the kavadi, women are more likely to exhibit signs of
divine possession by falling into trance, dancing uncontrollably, and
engaging in emotional displays during the procession. As trance is hard
to fake, this behavior may be perceived as an honest signal of devotion
(Power, 2017a; Soler, 2012). Indeed, people who regularly attend these
ceremonies often comment on the authenticity of performers' behavior,
and those who are perceived to be faking possession are denigrated.
Moreover, women also make signicant time investments throughout
the festival, including helping the pilgrims, assisting with preparations
at the temple, and hosting meals.
More generally, it is worth noting that many individuals participate
in different roles. From musicians and priests to family members who
accompany their beloved ones, various groups show their devotion to
the community in different ways. Additionally, various forms of
signaling may take place at the periphery of the ritual. One such example
is the ceremonial meal that concludes the festival. This meal, known as
the “seven curries”, is an elaborate dinner served by the families of
kavadi bearers at their home and attended by kin and kith. Prominent
members of the community sometimes host these feasts for dozens or
even hundreds of individuals. Similarly, it is common for those not
carrying a kavadi themselves to pour some water on the pilgrims' feet to
alleviate the pain caused by walking barefooted on the burning asphalt.
Afuent individuals may hire water trucks to get ahead of the procession
and pour tons of water on the road as it passes. Just like the act of
carrying a kavadi, these forms of investment too are sensitive to social
factors and can be used in different ways. Physical and nancial ex-
penditures can be used cumulatively to increase the strength of the
signal (Bliege Bird & Power, 2015), or in hydraulic fashion, as inter-
changeable forms of capital (Bourdieu, 1986). Future research may try
to examine those distinct forms of participation.
Our study focused on socioeconomic factors involved in different
forms of ritual signaling. It did not, however, attempt to measure the
potential benets that might be derived from this signaling, as our data
do not afford any conclusions about how participation actually affects
social status. We predict that those benets too will be relative to one's
Table 2
Untransformed estimates with 95% CI from negative binomial models (with log link) predicting the number of piercings during Kavadi.
All participants Males only
(1) (2) (3) (4) (5) (6)
Intercept 3.88 (3.47, 4.30) 0.87 (−0.48, 2.21) 0.9 (−0.47, 2.27) 4.31 (3.89, 4.74) 5.57 (3.72, 7.41) 5.54 (3.56, 7.51)
SES −0.33 (−0.75, 0.08) −0.26 (−0.61, 0.09) 0.08 (−0.38, 0.53) −0.43 (−0.85, −0.02) −0.33 (−0.77, 0.12) 0.09 (−0.53, 0.71)
Age – 0 (−0.04, 0.04) −0.002 (−0.04, 0.03) – 0.02 (−0.03, 0.07) 0.02 (−0.04, 0.07)
Sex – 4.14 (3.37, 4.91) 4.09 (3.32, 4.86) – – –
Marital status – −0.33 (−1.17, 0.51) −0.1 (−0.94, 0.75) – −0.52 (−1.71, 0.68) −0.12 (−1.36, 1.11)
Religiosity – −0.09 (−0.24, 0.06) −0.12 (−0.27, 0.03) – −0.15 (−0.37, 0.06) −0.21 (−0.43, 0.01)
Private ritual – −0.13 (−0.49, 0.24) 0.01 (−0.37, 0.38) – −0.16 (−0.60, 0.28) −0.003 (−0.45, 0.44)
Public ritual – 0.85 (−0.37, 2.07) −0.54 (−3.78, 2.70) – 0.89 (−0.53, 2.31) −0.15 (−3.96, 3.67)
SES*Public ritual – – −3.50 (−6.85, −0.14) – – −3.49 (−7.47, 0.50)
N 80 80 77 52 52 49
Note. SES is Socio-Economic Status; Age is centered at its mean; Reference category for Sex is female. Reference category for Martial status is unmarried.
Table 3
Untransformed estimates with 95% CI from gamma models (with log link) predicting kavadi size.
All participants Males only
(1) (2) (3) (4) (5) (6)
Intercept 0.25 (0.04, 0.45) −0.15 (−0.86, 0.56) −0.24 (−0.92, 0.44) 0.46 (0.24, 0.69) 0.18 (−0.77, 1.12) 0.27 (−0.63, 1.16)
SES 0.24 (0.04, 0.44) 0.12 (−0.06, 0.31) −0.14 (−0.37, 0.08) 0.19 (−0.02, 0.41) 0.23 (0.0005, 0.46) −0.04 (−0.32, 0.24)
Age – −0.03 (−0.05, −0.01) −0.03 (−0.04, −0.01) – −0.03 (−0.06, −0.01) −0.03 (−0.05, −0.005)
Sex – 0.91 (0.53, 1.29) 0.94 (0.58, 1.29) – – –
Marital status – 0.03 (−0.41, 0.47) −0.002 (−0.41, 0.41) – 0.34 (−0.27, 0.96) 0.27 (−0.29, 0.83)
Religiosity – −0.04 (−0.11, 0.04) −0.04 (−0.11, 0.03) – 0.01 (−0.10, 0.12) −0.01 (−0.11, 0.09)
Private ritual – −0.07 (−0.27, 0.13) −0.07 (−0.26, 0.12) – −0.02 (−0.24, 0.21) −0.04 (−0.24, 0.16)
Public ritual – 0.15 (−0.53, 0.83) 1.09 (−0.57, 2.75) – 0.24 (−0.49, 0.98) 1.47 (−0.26, 3.20)
SES*Public Ritual – – 2.13 (0.40, 3.86) – – 1.85 (0.05, 3.64)
N 80 80 77 52 52 49
Note. SES is Socio-Economic Status; Age is centered at its mean; Reference category for Sex is female; Reference category for Martial status is unmarried.
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
9
position in the social hierarchy (Shaver, 2015). That is, all else being
equal, high-status individuals who produce weaker signals (participate
in less intense forms of the kavadi) may reap the same benets as those
for low-status individuals who produce louder signals (practice more
intense forms of the rituals). Inversely, a signal of similar intensity will
result in higher benets for high-status individuals. For example, ob-
servers of the ritual may perceive a high-status person who performs the
ritual once as equally religious with a low-status person who performs it
thrice. On the other hand, a high-status individual with ten piercings
may be perceived as more religious than a low-status status individual
with the same number of piercings.
Our study quantied the impact of socio-economic factors on the
frequency and intensity of ritual signaling. Inevitably, however, our
measures are imperfect. Religiosity is notoriously hard to dene, let
alone to quantify (Xygalatas & Lang, 2016). Our ethnographic experi-
ence suggests that Mauritians do have a shared concept of the term
religion, and for that matter one that is comparable to Western con-
ceptions—besides, the word for religion in the local creole language is
the same Latin-origin word used in English. Still, a single-item question
cannot possibly capture the complexity of religious commitments, and
for that matter, neither can any multi-item one. Moreover, the frequency
of ritual practice was self-reported, and thus vulnerable to issues such as
social desirability, inaccurate estimates, or reliability of recall. These
limitations highlight the need for any quantitative studies of religion to
be anchored in systematic ethnographic observation, which can guide
researchers in asking culturally pertinent questions as well as in better
understanding and evaluating their ndings.
Collective events like the Thaipusam Kavadi are public arenas where
social hierarchies are established, renewed, and negotiated in myriad
ways (Bourdieu, 1977; Turner, 1979). In these contexts, the costs and
benets of each signal can and do vary between different subgroups such
as caste, class, and gender. For example, members of the upper classes
may often frown upon some of the most extreme piercings, while
members of the lower classes consider them acts of stronger devotion. It
seems that those who lack social capital are willing to pay a higher price
to achieve it, either in the form of nancial capital or (quite literally) in
the form of blood, sweat and tears. On the other hand, those who already
possess more social capital are rather interested in maintaining the
status quo. Indeed, high-ranking individuals are often in position to use
their status in order to manipulate ritual systems so as to decrease the
incentives for investment by those of lower status (Shaver, 2015).
For practical and ethical reasons, extreme rituals are not possible to
study under controlled conditions. Our design allowed us to examine
signaling aspects of the kavadi in a naturalistic setting. Rather than
relying on self-reports or hypothetical scenarios, we used behavioral
data in a real-life context. Our study was, to our knowledge, the rst to
quantify the multi-layered costs of an extreme ritual and to examine how
socioeconomic factors affect the differential distribution of these costs.
Our ndings provide evidence that is largely consistent with a costly
signaling view of ritual. Needless to say, however, cultural phenomena
are complex, and can never be reduced to monocausal explanations. The
current study has merely scratched the surface of one of the most widely
performed rituals in the world.
Acknowledgments
We thank Martin Kanovsky for his valuable comments. This study
was funded by the Interacting Minds Centre at Aarhus University,
Denmark, and the LEVYNA Laboratory for the Experimental Research of
Religion at Masaryk University, Czech Republic. DX acknowledges
support from the University of Connecticut Humanities Institute; ML and
RK from the Czech Science Foundation (GA CR) [18-18316S]; EKK from
the HUME Lab Experimental Humanities Laboratory, Faculty of Arts,
Masaryk University; PM from the Evolutionary and Cognitive Research
of Religion grant (MUNI/A/1444/2020) from Masaryk University; VB
and PM from the VEGA grant 2/0102/19, Collective Rituals as a tool of
social regulation; and JS from the Royal Society of New Zealand Marsden
Fund (19-UOO-090).
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.evolhumbehav.2021.05.006.
References
Barker, J., Power, E., Heap, S., Puurtinen, M., & Sosis, R. (2019). Content, cost, and
context: A framework for understanding human signaling systems. Evolutionary
Anthropology, 28(2), 86–99.
Belle, C. (2017). Thaipusam in Malaysia. Singapore: ISEAS.
Bliege Bird, R., & Power, E. A. (2015). Prosocial signaling and cooperation among Martu
hunters. Evolution and Human Behavior, 36(5), 389–397.
Bliege Bird, R., & Smith, E. A. (2005). Signaling theory, strategic interaction, and
symbolic capital. Current Anthropology, 46(2), 221–248.
Boone, J. (2000). Status signaling, social power, and lineage survival. In M. Diehl (Ed.),
Hierarchies in action: Qui bono? Carbondale, IL: Center for Archaeological Studies.
Bourdieu, P. (1977). Outline of a theory of practice. Trans. Richard Nice. Cambridge:
Cambridge University Press.
Bourdieu, P. (1986). The forms of capital. In J. G. Richardson (Ed.), Handbook of theory
and research for the sociology of education (pp. 241–258). New York: New York:
Greenwood Press.
Bulbulia, J. (2004). Religious costs as adaptations that signal altruistic intention.
Evolution and Cognition, 10(1), 19–38.
Bulbulia, J., Shaver, J. H., Greaves, L., Sosis, R., & Sibley, C. (2015). Religion and
parental cooperation: An empirical test of Slone’s sexual signaling model. In
J. Bulbulia, R. Sosis, E. Harris, R. Genet, & C. Genet (Eds.), The attraction of religion
(pp. 29–62). Santa Margarita: Bloomsbury Publishing.
Buser, T. (2015). The effect of income on religiousness. American Economic Journal:
Applied Economics, 7(3), 178–195.
Cronk, L. (1994a). Evolutionary theories of morality and the manipulative use of signals.
Zygon, 29(1), 81–101.
Cronk, L. (1994b). The use of moralistic statements in social manipulation: A reply to
Roy A. Rappaport. Zygon, 29(1), 351–355.
Cronk, L. (2005). The application of animal signaling theory to human phenomena: Some
thoughts and clarications. Biology and Social Life, 4(44), 603–620.
Darwin, C. (1859). On the origin of species by means of natural selection. London: John
Murray.
Darwin, C., & Wallace, A. (1858). On the tendency of species to form varieties; and on the
perpetuation of varieties and species by natural means of selection. Journal of the
Proceedings of the Linnean Society of London. Zoology, 3(9), 45–62.
Fischer, R., Xygalatas, D., Mitkidis, P., Reddish, P., Konvalinka, I., & Bulbulia, J. (2014).
The re-walker’s high: Affect and physiological responses in an extreme collective
ritual. PLOS ONE, 9(2), Article e88355.
Grafen, A. (1990). Biological signals as handicaps. Journal of Theoretical Biology, 144(4),
517–546.
Grose, J. (2011). Modelling and the fall and rise of the handicap principle. Biology and
Philosophy, 26(5), 677–696.
Guilford, T., & Dawkins, M. (1991). Receiver psychology and the evolution of animal
signals. Animal Behaviour, 42(1), 1–14.
Henrich, J. (2009). The evolution of costly displays, cooperation and religion: Credibility
enhancing displays and their implications for cultural evolution. Evolution and
Human Behavior, 30(4), 244–260.
Iannaccone, L. (1992). Sacrice and stigma: Reducing free-riding in cults, communes,
and other collectives. Journal of Political Economy, 100(2), 271–291.
Iannaccone, L. (1994). Why strict churches are strong. American Journal of Sociology,
1180–1211.
Iannaccone, L. (1998). Introduction to the economics of religion. Journal of Economic
Literature, 36(3), 1465–1495.
Irons, W. (2001). Religion as a hard-to-fake sign of commitment. In R. M. Nesse (Ed.),
Evolution and the capacity for commitment (pp. 292–309). New York: Russell Sage
Foundation.
Leach, E. (1954). Political systems of highland Burma: A study of Kachin social structure.
Cambridge, MA: Harvard University Press.
Maynard Smith, J., & Harper, D. (2003). Animal signals. Oxford: Oxford University Press.
Partan, S., & Marler, P. (1999). Communication goes multimodal. Science, 283(5406),
1272–1273.
Pendall, R., Hayes, C., George, A., McDade, Z., Dawkins, C., Jeon, J., … Smart, M.
(2014). Driving to opportunity: Understanding the links among transportation access,
residential outcomes, and economic opportunity for housing voucher recipients.
Wasington DC: Urban Institute.
Power, E. A. (2017a). Discerning devotion: Testing the signaling theory of religion.
Evolution and Human Behavior, 38(1), 82–91.
Power, E. A. (2017b). Social support networks and religiosity in rural South India. Nature
Human Behavior, 1(3), 1–6.
Power, E. (2018). Collective ritual and social support networks in rural South India.
Proceedings of the Royal Society B, 285(1879), 20180023.
R Core Team. (2019). R: A language and environment for statistical computing. Vienna: R
Foundation for Statistical Computing.
D. Xygalatas et al.
Evolution and Human Behavior xxx (xxxx) xxx
10
Rowe, C. (1999). Receiver psychology and the evolution of multicomponent signals.
Animal Behaviour, 58, 921–931.
Sauzier-Uchida, E. (2009). Language choice in multilingual Mauritius: National unity
and socioeconomic advancement. Journal or Liberal Arts, 126, 99–130.
Shaver, J. H. (2015). The evolution of stratication in Fijian ritual participation. Religion,
Brain & Behavior, 5(2), 101–117.
Shaver, J. H., & Sosis, R. (2014). How does male ritual behavior vary across the lifespan?
Human Nature, 25(1), 136–160.
Shaver, J. H., & Sosis, R. (2018). Costly signaling in human cultures. In C. Power (Ed.),
International encyclopedia of anthropology: Evolutionary and biosocial perspectives in
anthropology. Wiley-Blackwell.
Shavers, V. L. (2007). Measurement of socioeconomic status in health disparities
research. Journal of the National Medical Association, 99(9), 1013.
Smith, E. A., & Bliege Bird, R. (2000). Turtle hunting and tombstone opening: Public
generosity as costly signaling. Evolution and Human Behavior, 21, 245–262.
Smith, E. A., Bliege Bird, R., & Bird, D. W. (2003). The benets of costly signaling:
Meriam turtle hunters. Behavioral Ecology, 14(1), 116–126.
Soler, M. (2012). Costly signaling, ritual and cooperation: Evidence from Candombl´
e, an
Afro-Brazilian religion. Evolution and Human Behavior, 33(4), 346–356.
Soler, M., Batiste, F., & Cronk, L. (2014). In the eye (and ears) of the beholder: Receiver
psychology and human signal design. Evolutionary Anthropology, 23(4), 136–145.
Sosis, R. (2000). Costly signaling and torch shing on Ifaluk Atoll. Evolution and Human
Behavior, 21(4), 223–244.
Sosis, R. (2003). Why aren’t we all Hutterites? Human Nature, 14(2), 91–127.
Sosis, R., & Alcorta, C. (2003). Signaling, solidarity, and the sacred: The evolution of
religious behavior. Evolutionary Anthropology, 12(6), 264–274.
Sosis, R., & Bressler, E. (2003). Cooperation and commune longevity: A test of the costly
signaling theory of religion. Cross-Cultural Research, 37(2), 211.
Sosis, R., & Rufe, B. (2003). Religious ritual and cooperation: Testing for a relationship
on Israeli religious and secular kibbutzim. Current Anthropology, 44(5), 713–722.
Statistics Mauritius. (2012). 2011 housing and population census, Republic of Mauritius:
Demographics and fertility characteristics. Port Louis.
Turner, V. (1979). Frame, ow and reection: Ritual and Drama as public Liminality.
Japanese Journal of Religious Studies, 6(4), 465–499.
Van Buuren, S., & Groothuis-Oudshoorn, K. (2011). Multivariate imputation by chained
equations. Journal of Statistical Software, 45(3), 1–67.
Vyas, S., & Kumaranayake, L. (2006). Constructing socio-economic status indices: How to
use principal components analysis. Health Policy and Planning, 21(6), 459–468.
Ward, C. M. (1984). Thaipusam in Malaysia: A psycho-anthropological analysis of ritual
trance, ceremonial possession and self-mortication practices. Ethos, 12(4),
307–334.
Wickham, H. (2016). ggplot2: Elegant graphics for data analysis. New York: Springer-
Verlag.
Wiley, R. H. (2015). Noise matters. The evolution of communication. Cambridge, MA:
Harvard University Press.
Xygalatas, D. (2012). The burning saints. London: Equinox.
Xygalatas, D. (2013). Effects of religious setting on cooperative behaviour. A case study
from Mauritius. Religion, Brain and Behavior, 3(2), 91–102.
Xygalatas, D. (2015). The biosocial basis of collective effervescence: An experimental
anthropological study of a re-walking ritual. Fieldwork in Religion, 9(1), 53–67.
Xygalatas, D., & Lang, M. (2016). Prosociality and religion. In N. Kasumi Clements (Ed.),
Mental religion. Macmillan interdisciplinary handbooks (pp. 119–133). Macmillan.
Xygalatas, D., & Maˇ
no, P. (2021). Ritual exegesis among Mauritian Hindus.
Xygalatas, D., Mitkidis, P., Fischer, R., Reddish, P., Skewes, J., Geertz, A. W., et al.
(2013). Extreme rituals promote prosociality. Psychological Science, 24(8),
1602–1605.
Xygalatas, D., Kundtov´
a Klocov´
a, W., Cig´
an, J., Kundt, R., Maˇ
no, P., Kotherov´
a, S., …
Kanovsky, M. (2016). Location, location, location: Effects of cross-religious primes
on prosocial behaviour. International Journal for the Psychology of Religion, 26(4),
304–319.
Xygalatas, D., Kotherov´
a, S., Maˇ
no, P., Kundt, R., Cig´
an, J., Kundtov´
a Klocov´
a, E., &
Lang, M. (2017). Big gods in small places: The random allocation game in Mauritius.
Religion, Brain and Behavior, 8(2), 243–261.
Xygalatas. (2021). The power of ritual. Prole Books. forthcoming.
Xygalatas, D., Khan, S., Lang, M., Kundt, R., Kundtov´
a-Klocov´
a, E., Kratky, J., &
Shaver, J. (2019). Effects of extreme ritual practices on health and well-being.
Current Anthropology, 60(5), 699–707.
Zahavi, A. (1975). Mate selection—A selection for a handicap. Journal of Theoretical
Biology, 53(1), 205–214.
Zahavi, A., & Zahavi, A. (1997). The handicap principle: A missing piece of Darwin’s puzzle.
New York: Oxford University Press.
D. Xygalatas et al.
