Content uploaded by Thomas J. Coleman III
Author content
All content in this area was uploaded by Thomas J. Coleman III on Aug 02, 2018
Content may be subject to copyright.
Content uploaded by Thomas J. Coleman III
Author content
All content in this area was uploaded by Thomas J. Coleman III on Aug 02, 2018
Content may be subject to copyright.
Content uploaded by Thomas J. Coleman III
Author content
All content in this area was uploaded by Thomas J. Coleman III on Jul 18, 2018
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=rrbb20
Religion, Brain & Behavior
ISSN: 2153-599X (Print) 2153-5981 (Online) Journal homepage: http://www.tandfonline.com/loi/rrbb20
Predicting age of atheism: credibility enhancing
displays and religious importance, choice, and
conflict in family of upbringing
Joseph Langston, David Speed & Thomas J. Coleman III
To cite this article: Joseph Langston, David Speed & Thomas J. Coleman III (2018): Predicting
age of atheism: credibility enhancing displays and religious importance, choice, and conflict in
family of upbringing, Religion, Brain & Behavior, DOI: 10.1080/2153599X.2018.1502678
To link to this article: https://doi.org/10.1080/2153599X.2018.1502678
Published online: 30 Jul 2018.
Submit your article to this journal
Article views: 25
View Crossmark data
Predicting age of atheism: credibility enhancing displays and
religious importance, choice, and conflict in family of upbringing
Joseph Langston
a
, David Speed
b
and Thomas J. Coleman III
c
a
Atheist Research Collaborative, Colorado Springs, CO, USA;
b
Department of Psychology, University of New
Brunswick, Saint John, NB, Canada;
c
Centre for Advances in Behavioural Science; Brain, Belief, and Behaviour
Research Lab, Coventry University, Coventry, UK
ABSTRACT
The cultural learning concept of Credibility Enhancing Displays (CREDs)
concerns the extent to which behavioural models consistently live out
their professed ideals. While researchers have suggested that past CRED
exposure is an important variable for predicting who does and does not
become a religious believer, it is unclear how CREDs relate to when a
person rejects the religious beliefs modelled to them during their
upbringing. Using a large sample of formerly believing atheists, two
analyses assessed the ability of CREDs to predict the age at which an
individual became an atheist. In the first analysis (n= 5,153), CREDs were
positively associated with a delay in Age of Atheism, with family-level
religious variables (Religious Importance, Religious Choice, and Religious
Conflict) moderating this relationship. In the second analysis (n= 3,210),
CREDs remained a stable predictor of Age of Atheism while controlling
for demographics, parental quality, religious variables, relational
variables, and institutional variables. Overall, while findings support a
robust relation of CREDs to atheistic outcomes even when controlling
for many other variables that influence religious transmission processes,
they also highlight the importance of considering how such other
variables modify the impact of CREDs on (non)religious outcomes.
ARTICLE HISTORY
Received 28 May 2018
Accepted 15 July 2018
KEYWORDS
Credibility enhancing
displays; atheism; religious
conflict; religious choice;
parental quality; religious
socialization; cognitive
science of religion
1. Introduction
In order to understand the ubiquity of specific concepts and behaviours that have been referred to as
“religious”, the cognitive science of religion (CSR) focuses upon recurring cognitive and cultural pro-
cesses that have evolved or otherwise developed across the world and over time (Xygalatas, 2014).
Over the last few decades, researchers in this area have proposed a number of cognitive and cultural
constructs linked to varying levels of religious beliefs and ideas (cf. Barrett, 2007). Because CSR is still
a relatively new area of research, many open questions remain concerning these theoretical con-
structs, particularly how they relate to religious conditions and outcomes across different cultures
in the modern social landscape. One of these constructs is credibility enhancing displays (CREDs),
introduced in the work of anthropologist Henrich (2009). Derived from a cultural evolution stand-
point, CREDs imply a “walk the walk”notion: if individuals faithfully engage in the behaviours that
are logical expressions of their own professed beliefs, then this should increase the probability that
those who observe these behaviours will adopt the beliefs that underlie such behaviours. More pre-
cisely, CREDs are behaviours that a cultural model would not perform if they did not believe what
© 2018 Informa UK Limited, trading as Taylor & Francis Group
CONTACT Joseph Langston staff@atheistresearch.org, langston1029@gmail.com
RELIGION, BRAIN & BEHAVIOR
https://doi.org/10.1080/2153599X.2018.1502678
they said they did. As Gervais and colleagues (Gervais, Willard, Norenzayan, & Henrich, 2011,
p. 392) explain:
An unscrupulous [behavioural] model might knowingly transmit false information to others, perhaps to main-
tain a competitive advantage. In this case, it is important for learners to ensure that their models actually hold
the beliefs they espouse before adopting the belief themselves. Actions speak louder than words: they are great
cues of another’s underlying beliefs …If models engage in behaviours that would be costly if opposing beliefs
were held (that is, if they engage in credibility-enhancing displays of their beliefs) learners can be more
confident that the model actually holds the belief, and as a result they would be more receptive to these beliefs
(Henrich, 2009).
While CREDs are related to various CSR constructs, such displays are perhaps best described
as constituting a cultural evolution construct, insofar as CREDs were partially developed in
response to, and as a critique of, CSR emphases on content biases rather than context biases
(see Gervais & Henrich, 2010, for a discussion of context versus content biases). As such, the
importance of CREDs can be understood not only in relation to the physical survival and genetic
reproduction of humans via the learning of which behaviours to perform (i.e., obtaining food;
cultural norms) or not perform (i.e., avoiding dangerous areas, foods, and animals), but also
specifically in relation to social learning, in this case as it pertains to religious beliefs and beha-
viours (Bandura, 1971,2003; cf. also Dudley, 1999). Stated bluntly, we have an evolved bias
toward the learning of behaviours that keep us alive and enhance or ensure successful reproduc-
tion, but this bias also pertains to how and why we learn the various aspects of our culture from
important cultural models. Consequently, since religious socialization is central to the interge-
nerational transmission of religious beliefs and behaviours (Bengtson, Putney, & Harris, 2013),
evolved cultural learning mechanisms that bias humans toward conforming to the behaviour
of important cultural models should be more influential than verbal expressions of religious
beliefs from such models. While research shows that either of these aspects may influence reli-
gious outcomes among offspring (Flor & Knapp, 2001; King, Furrow, & Roth, 2002), CREDs
should play a unique role in the youthful acquisition and later sustainment of religious ideas
and practices.
Although they are not entirely without mixed results, a number of studies provide support for the
efficacy of CREDs in distinguishing between believers (i.e., theists) and nonbelievers (i.e., atheists), as
well as general levels of religiosity in national populations (Gervais & Najle, 2015; Hitzeman &
Wastell, 2017; Lanman, 2010,2012; Lanman & Buhrmester, 2016; Maij et al., 2017; Willard &
Cingl, 2017).
1
Because such research suggests that CREDs influence the acquisition and intergenera-
tional transmission of religious belief, a logical implication of this is that CREDs should impact the
age at which one becomes an atheist. However, CREDs do not exist within a vacuum, but are likely
related to other important influences on the acquisition of (religious) culture (e.g., family-level vari-
ables). That is, if CREDs are related to the age at which one stops believing in a god or gods, then
there is a question of how CREDs might fare next to, interact with, or even compete with other social
and family variables which have been shown to affect religious outcomes. Such variables would
include (among many others; cf. Clark & Worthington, 1990): (1) peer religiosity and religiosity
of social networks (Barry & Christofferson, 2014); (2) family structure (e.g. two-parent versus one
parent) (Petts, 2014); (3) parental religious homogamy (Bruce & Glendinning, 2010); (4) quality
of relationship with parents and other key social relationships (e.g., grandparents, siblings, adoles-
cent and young adulthood peers; Smith & Denton, 2005); (5) degree of personal religious choice
allowed to children by parents (Okagaki, Hammond, & Seamon, 1999; Potvin & Sloane, 1985);
(6) conflict over religion in one’s family during upbringing (Bengtson et al., 2013, pp. 160–161;
Mahoney, 2005; Pasquale, 2009); and (7) formal religious education and institutional religious
contexts (Putney et al., 2013).
Of these factors, religious choice and religious conflict are of special interest. The former may
especially be considered in post-industrial, modernized liberal democracies (Norris & Inglehart,
2011), which, purportedly due to their increased levels of existential security, are generally
2J. LANGSTON ET AL.
characterized by elevated levels of personal choice, individualism, and secular liberal values (i.e., per-
sonal autonomy; Thiessen & Wilkins-Laflamme, 2017; Voas & Doebler, 2011,2014; cf. also Berkers,
2018, in press). Lanman (2012) has suggested that increased levels of existential security should lower
the overall number of CREDs in the cultural learning environment, thus rendering belief in super-
natural agents less plausible for future generations. At any rate, insofar as the ability and effort of
parents to transmit religious values and ideologies is compromised by such environments (whether
through fewer CREDs or greater personal choice), this could influence a population increase in athe-
ism. In line with this, Voas and Doebler (2014) suggested that religious change across generations is
closely connected to the relation between religion and childrearing values: parents may have become
less committed than in the past to ensuring religious conformity in their children, and thus perhaps
more committed to allowing their children to make their own choices about religious matters (for
supporting arguments, see also Thiessen, 2016; Manning, 2015, on nonreligious parenting and values
of choice and autonomy).
Religious choice may have an expected relationship with conflict over religion in the family. If
children and adolescents in the aforementioned cultural contexts are subject to greater expectations
of personal choice and self-determination, then such an expectation, when upset by (religious)
authoritarian parenting, might lead to conflict over religion and subsequently result in degrees of
alienation, personal disappointment, and rebellion, regarding both religion and one’s parents.
2
Pas-
quale (2009, p. 82) suggested that “more aggressive or stringent [religious] doctrinal or behavioural
expectations will tend, on average, to produce greater numbers of individuals who experience con-
fusion, conflict, or critical reaction.”Similarly, in discussing families of nonreligious youth, Bengtson
et al. (2013, pp. 160–161) explicitly stated that family religious conflict has been identified in other
research as a common path to atheism. As they pointed out, the religious socialization efforts of
overly demanding and zealous parents can work against the goal of transmitting their religious tra-
dition to offspring. In describing such conflict with parents over religion, tellingly, one nonreligious
respondent’s childhood was described as not allowing for individual choice in matters of religion;
this person went on to become an atheist.
Defining atheism explicitly as non-belief in a god or gods (i.e., a lack of theism; cf. also Cliteur,
2009), and using a sample consisting only of former believers, in the current paper we explored
the relationship between CREDs and the age of atheism using two main analyses. First, we sought
to determine if CREDs predicted age of atheism, and whether this relationship would be influenced
by three specific family- and parent-level religious variables (i.e., religious importance, religious
choice, and religious conflict). As part of this analysis, we also tested statistical interactions between
CREDs and religious importance, religious choice, and religious conflict.
Second, we investigated whether the relationship between CREDs and age of atheism was sub-
stantially attenuated or eliminated by the inclusion of a broader assortment of other social and family
covariates (as enumerated above) which have been shown to affect the acquisition and transmission
of religious beliefs. Because CREDs are related to broader (religious) socialization, there is a question
of the relationship to CREDs of other variables which are regarded as mainstays of the religious
socialization process. While we would expect higher levels of parental religious CREDs to be associ-
ated with a delay in the “onset”of atheism, little is known about the influence of other candidate
family, social, and religion variables in relation to CREDs, regarding their collective impact on
age of atheism.
2. Methods
2.1. Participants
Data were collected via an online Qualtrics survey from September to October in 2017. The study
was advertised as research into how family and social processes influence becoming an atheist.
We explicitly stated that a person was not eligible for the study if they (a) currently believed in
RELIGION, BRAIN & BEHAVIOR 3
the existence of god(s), or if they (b) believed in a spiritual or higher power. We included two general
instructions prior to the survey, quoted as follows:
(1) In the questions and answers below, we use the words “god”and “gods”. Please interpret these
terms to stand for whatever image or idea you primarily associate with them, such as a specific
god or gods you once believed existed, or a specific god or gods that other people believe exist(s).
(2) At least one existing research study indicates that people who call themselves “atheists”do not
uniformly agree on the definition of the term. Other research indicates that not every person
who either (a) believes that god or gods do not exist or (b) lacks a belief in god or gods, self-
identifies or uses the label “atheist”for themselves. However, for the purposes of this current
study, we use the term “nonbeliever”and we refer broadly to “not believing in”the existence
of god or gods (see #1, above).
We sent our survey to more than 100 atheist, secular, and freethought organizations across a glo-
bal setting, to include social media platforms (e.g., Facebook, Reddit). After removing incomplete
cases and those who incorrectly answered any one of three “attention check”questions, a total of
7,173 respondents remained. In the current study we report on a maximum total of 5,153 respon-
dents who answered all questions of interest. Because our second main analysis included many
additional variables, we provide descriptive statistics for both samples in Table 1, as they were
used in each analysis.
A special note of interpretive caution regarding our sample must be sounded before we can con-
tinue. As one of our reviewers pointed out, our sample was largely acquired via organizational chan-
nels, and thus likely does not represent the growing collection of nonbelievers in Western societies
more generally. In other words, they would be highly atypical outside of contexts and countries
where “religion”is not (or is no longer) a salient part of history and culture (Garcia & Blankholm,
2016; Zuckerman, 2012). As a result, our sample is likely mostly constituted by persons for whom
religious nonbelief is an important component of their social identities, as they were ultimately
self-selected into a study targeted at a specific subcultural group within the broader atheist milieu.
Consequently, we are probably not studying the kind of nonbelievers that would turn up in a nation-
ally representative probability sample. In this sense, our analysis would most properly be regarded as
centreing upon, not generic atheists or atheism per se, but on the explicit rejection of prior religious
beliefs, or self-aware atheist converts, and the age at which they actively rejected the religion of their
upbringing.
2.2. Measures
2.2.1. Age of atheism
The dependent variable was captured with the question, “At what age did you no longer believe in
god(s)?”. Because we sought to recruit only those who had once been believers in a god or gods but
were no longer, we chose to exclude from analyses those respondents who reported becoming an
atheist prior to the age of five.
2.2.2. Demographic covariates
Sex (ref = female), current age, race (ref = non-white), education (ref = high school or less), marital
status (ref = single/never married), religious upbringing (ref = None) and country of residence (ref =
United States; four other categories included Canada, Great Britain, Australia, and Other) were all
included as covariates in predictive analyses. Due to the fact that the “Other”country category
was large (75 countries in total), its meaningfulness as an analytical category in our models should
be considered limited.
4J. LANGSTON ET AL.
Table 1. Descriptive statistics for Model 1 and Model 2.
Model 1
(n= 5,153)
Model 2
(n= 3,210)
M (SD) % M (SD) %
Age of Atheism 23.8 (10.8) 24.1 (10.8)
Sex
Other 1.5 1.5
Male 43.2 42.3
Female 55.4 56.2
Age 41.3 (14.8) 41.2 (14.9)
Ethnicity
White 91.0 91.0
Non-white 9.0 9.0
Education
High School 6.5 6.6
Some Post-secondary 19.7 19.3
Post-Secondary 42.7 43.0
Graduate School 26.8 26.7
Other 4.2 4.4
Marital Status
Single 18.7 18.3
Married/Common-Law 67.8 68.0
Widowed/Separated/Div. 10.8 11.1
Other 2.8 2.7
Religious Upbringing
None 2.4 2.2
Christian 90.0 90.6
Jewish 1.6 1.5
Muslim 0.85 0.84
Other 5.1 5.0
Country
United States 86.4 86.6
Canada 4.4 4.4
United Kingdom 1.3 1.4
Australia 1.7 1.4
Other 6.3 6.1
Religious Importance 5.2 (2.0) 5.3 (2.0)
Religious Choice 3.5 (2.4) 3.4 (2.3)
Religious Conflict 3.5 (2.2) 3.5 (2.2)
CREDs 26.3 (10.8) 26.8 (10.8)
Mother Strict 4.4 (1.6)
Mother Warm 5.0 (1.8)
Mother Talk 4.1 (1.9)
Dad Strict 4.6 (1.8)
Dad Warm 3.9 (1.8)
Dad Talk 3.4 (1.9)
Parental Relationship
Lived with parents 76.2
Did not live with parents 23.8
Parental Religion
Parents different relig. 17.2
Parents same relig. 82.8
Non-Believing Friends
Knew other atheists 29.7
Did not know atheists 70.3
Peer Religiosity Index 1 8.9 (6.1)
Peer Religiosity Index 2 9.2 (5.4)
Worship service attend
No 56.3
Yes 37.4
Stopped before age 13 6.3
Years religious school 3.4 (4.9)
RELIGION, BRAIN & BEHAVIOR 5
2.2.3. Quality of parental relationships
Six questions (cf. Hunsberger, Pratt, & Pancer, 2002) assessed parental quality: “While you were
growing up, would you say that your [Mother or Father] was (1) Easy to Talk With, (2) Strict,
and (3) Warm and Loving.”Each of these was rated on a seven-point Likert scale (1 = Not at all,
4 = Somewhat, 7 = Completely). In subsequent analyses, Strictness ratings were reverse coded.
2.2.4. Family religion variables
Three items addressed various aspects of religious dynamics in one’s family of upbringing, including
Religious Importance (“Religion was important in my family.”), Religious Choice (“My parent(s)/
caretaker(s) allowed me to make my own choices about religious beliefs and practices.”), and Reli-
gious Conflict (“There was conflict between my parent(s)/caretaker(s) and I about religion.”). Each
item was rated on a seven-point Likert scale (1 = Strongly disagree, 4 = Neither agree nor disagree, 7
= Strongly agree).
2.2.5. Credibility enhancing displays (CREDs)
We used Lanman and Buhrmester’s(2016) seven-item CREDs measure (see appendix), which cap-
tures respondents’perceptions and self-reports of a variety of parental and/or caregiver religious
behaviours during upbringing. Each item was rated on a seven-point Likert scale (1 = To no extent
at all, 7 = To an extreme extent). Consistent with previous validation by the scale’s creators, principal
components analysis and factor analysis (principal axis factoring, Varimax rotation) revealed that all
seven items loaded onto a single factor (Eigenvalue = 4.3, 61.2%); loadings ranged from .70 to .86.
Therefore, we indexed all seven items (Cronbach’sα= .91), with higher scores indicating greater
CREDs exposure during upbringing.
2.2.6. Peer religiosity
We employed eight questions from a recently developed measure of peer religiosity (cf. Tratner et al.,
2017). For these items, we asked respondents to think of the one person who they would have con-
sidered to be their best friend for most of their adolescent years (ages 13 to 18). Three questions refer-
enced frequency of certain behaviours (1 = never;7=very often), e.g. “We prayed together.”The
remaining questions were answered by degree of true or untrue (1 = very untrue;7=very true),
e.g. “My friend showed their faith by how they talked and acted.”Initial principal components analy-
sis and factor analysis (principal axis factoring, oblimin rotation) revealed a two-factor solution
across the eight items. Based upon weak factor loadings and inter-item correlations, we dropped
one item (“We agreed in our religious attitudes and beliefs”). A subsequent promax rotation yielded
a more simplified factor structure than the oblimin rotation, and retained a two-factor solution, with
the two factors correlated at .57. The first 4-item factor was interpreted as the extent to which one
shared religious activities with one’sbest friend, whereas the second 3-item factor was interpreted as
synonymous with CREDs; that is, it reflected the fidelity and consistency of peer behavioural reli-
gious modelling.
2.2.7. Family structure and parental religious homogamy
One question addressed whether respondents grew up with (or, while growing up, mostly lived with)
both biological parents (dummy coded, 1 = yes). A second question asked whether respondents’
parents shared the same religion while the respondent was growing up (dummy coded, 1 = yes).
2.2.8. Known atheist others
One question addressed whether or not the respondent knew any other atheist(s) prior to themselves
becoming an atheist (dummy coded, 1 = yes), regardless of whether this was a friend, family member,
acquaintance, or significant other.
6J. LANGSTON ET AL.
2.2.9. Institutional religiosity
One item addressed whether respondents had incurred a two-year gap in religious worship service
attendance between the ages of 13 and 18 (Yes, No, Never/hardly ever attended before age 13).A
second item addressed the number of years an individual attended religious school during upbring-
ing (0 = Never attended religious school, 1 = Attended 1 year of religious school, 2 = Attended 2 years
of religious school, etc.)
2.3. Procedure and analysis plan
The current study had two main analyses, both of which explored the relationship between CREDs
and Age of Atheism, but using different variables of interest. The major thrust of all analyses was to
determine how different social factors influenced a person’s age of arrival at nonbelief in a god or
gods, in relation to CREDs. We tested underlying statistical assumptions for the analyses used,
and made corrections where appropriate (e.g., correcting standard error terms due to
heteroscedasticity).
2.3.1. Approach for the first set of analyses
The purpose of the first analysis (Table 2) was twofold: first, to determine the relationship between
CREDs, Religious Importance, Religious Choice, Religious Conflict, and Age of Atheism; and
second, to determine if this relationship was moderated by family-religion variables (importance,
choice, and conflict). We used hierarchical linear regression to determine if CREDs and our three
family religion variables predicted Age of Atheism, while controlling for demographic variables.
Because Breusch–Pagan tests were significant for heteroscedasticity, χ
2
= 1111.00, p < .001, we
used robust standard errors (i.e., HC3 corrections) with coefficients. Age of Atheism, Religious
Importance, Religious Choice, Religious Conflict, and CREDs were all standardized (M=0, SD =
1) to improve the interpretability of the overall models (West, Aiken, & Krull, 1996). As such, the
reported regression coefficients for Age of Atheism, Religious Importance, Religious Choice, and
Religious Conflict could be interpreted as changes in their respective standard deviations. For
example, in Block 2 of Table 2, CREDs has a B/ß coefficient of 0.14. This means that for every 1
unit increase in standard deviation for CREDs, one would expect an increase of 14% of 1 unit of stan-
dard deviation for Age of Atheism. Seeing that Age of Atheism had a standard deviation of 10.4 years
(Table 1), the reader could infer that increasing CREDs one unit of standard deviation was associated
with a delay in the Age of Atheism by approximately 1.5 years (10.4*0.14 = 1.456).
After we determined if CREDs, Religious Importance, Religious Choice, and Religious Conflict
predicted Age of Atheism, we then explored moderation terms. Specifically, we investigated if
CREDs’prediction of Age of Atheism changed as a function of Religious Importance, Religious
Choice, and Religious Conflict. That is, we investigated whether CREDs became more or less impor-
tant in predicting Age of Atheism when controlling for whether a respondent had high, moderate, or
low levels of Religious Importance, Religious Choice, and Religious Conflict. When or if moderation
terms for CREDs and family religion variables were significant, meaning that CREDs’importance
changed as a function of Religious Importance, Religious Choice, and Religious Conflict, we then
investigated how groups differed from each other using marginal means tests.
2.3.2. Approach for the second set of analyses
We conducted a second analysis (see Table 3) to explore how CREDs predicted Age of Atheism, and
to determine how this relationship was impacted by (religious) socialization variables that we
selected from the research literature explicitly because they have been shown to influence the acqui-
sition and transmission of religious beliefs and identity. This was accomplished through serial entry
and removal of model blocks. As can be seen in Table 3, CREDs remained in the hierarchical linear
regression model for entire set of analyses, but its coefficients fluctuate as a product of being entered
into the block along with other candidate variables.
RELIGION, BRAIN & BEHAVIOR 7
Table 2. Relationship between CREDs and age of atheism moderated by religious importance, choice, and conflict (n= 5,153).
Coefficients/Robust Standard Error
Block 1 Block 2 Block 3 Block 4 Block 5 Block 6
Constant −1.31 (.12)*** −1.07 (.12)*** −.93 (.12)*** −1.10 (.13)*** −.99 (.13)*** −.92 (.12)***
Ref = Female
Other .08 (.12) .07 (.11) .07 (.11) .07 (.11) .08 (.11) .08 (.11)
Male −.05 (.03) −.08 (.03)** −.10 (.03)** −.10 (.03)** −.10 (.03)** −.10 (.03)**
Age .03 (.001)*** .02 (.001)*** .03 (.001)*** .03 (.001)*** .03 (.001)*** .03 (.001)***
Ethnicity −.01 (.05) −.02 (.05) −.03 (.05) −.04 (.05) −.03 (.05) −.03 (.05)
Ref = High School
Some post-sec .05 (.06) .03 (.06) .03 (.06) .02 (.06) .03 (.06) .03 (.06)
Post-sec .16 (.05)** .13 (.05) * .11 (.05) * .11 (.05) * .11 (.05) * .11 (.05) *
Grad school .07 (.06) .04 (.06) .02 (.06) .01 (.06) .01 (.06) .01 (.06)
Other .16 (.10) .15 (.10) .13 (.09) .12 (.09) .13 (.09) .13 (.09)
Ref = Single
Married .06 (.03) .07 (.03) * .05 (.03) .05 (.03) .05 (.03) .05 (.03)
Wid./Sep./Div .06 (.06) .07 (.06) .06 (.06) .05 (.06) .06 (.06) .05 (.06)
Other .08 (.09) .10 (.09) .07 (.09) .06 (.09) .07 (.09) .07 (.09)
Ref = None
Christian .14 (.09) −.08 (.10) −.13 (.10) −.01 (.11) −.08 (.10) −.14 (.10)
Jewish −.41 (.15)** −.55 (.15)*** −.58 (.16)*** −.46 (.16)** −.53 (.16)** −.60 (.16)***
Muslim .27 (.13) * .01 (.13) −.06(.14) .06 (.14) −.02 (.14) −.07 (.14)
Other .30 (.11)** .04 (.11) .01 (.12) .11 (.12) .04 (.12) −.01 (.12)
Ref = USA
Canada .03 (.07) .03 (.07) .05 (.07) .05 (.07) .04 (.07) .05 (.07)
United Kingdom −.51 (.13)*** −.44 (.12)*** −.43 (.12)*** −.47 (.12)*** −.44 (.12)*** −.41 (.12)**
Australia −.15 (.10) −.11 (.10) −.10 (.10) −.11 (.10) −.10 (.10) −.09 (.10)
Other −.08 (.05) −.05 (.05) −.03 (.05) −.03 (.05) −.03 (.05) −.03 (.05)
CREDs (ß) .14 (.02)*** .06 (.02)** .04 (.02) .06 (.02)** .06 (.02)**
Rel. importance (ß) .11 (.02)*** .16 (.03)*** .12 (.02)*** .11 (.02)***
Rel. choice (ß) −.06 (.02)** −.06 (.02)** −.06 (.02)** −.06 (.02)**
Rel. conflict (ß) −.20 (.01)*** −.19 (.01)*** −.20 (.01)*** −.20 (.01)***
Importance*CREDs .08 (.02)***
Choice*CREDs −.03 (.02) *
Conflict*CREDs −.03 (.02) *
R
2
/ΔR
2
.16 *** .18/.02 *** .22/.04 *** .22/.01 *** .22/.01 * .22/.01 *
Note: We standardized select coefficients (ß) to aid with the interpretation of the overall model. This did not change the significance or meaning of the underlying model. For variables with ß, the
coefficient indicates the change in the DV (in units of SD). For example, moving from 0 to 1 on “CREDs”, which is the equivalent of moving 1 SD unit on “CREDs”, is associated with a .140 SD increase in
Age of Atheism in Block 2. Also note, ΔR
2
values for Block 4, Block 5, and Block 6, are giving an indication of change relative to Block 3.
*p<.05, ** p< .01, *** p< .001.
8J. LANGSTON ET AL.
Table 3. Relationship for CREDs predicting age of atheism while controlling for constructs on a rotating basis (n= 3,210).
Block 1 Block 2 Block 3 Block 4 Block 5 Block 6 Block 7
CREDs Only Demo. Variables Parent Quality Religious Variables Relational Variables Institution Variables Complete Model
Constant 19.9 (.51)*** 7.2 (1.5)*** 21.2 (1.1) 24.9 (1.04)*** 19.3 (.56)*** 24.4 (.69)*** 14.2 (1.95)***
CREDs .16 (.02)*** .15 (.02)*** .18 (.02) *** .09 (.02)*** .15 (.02)*** .07 (.02)** .04 (.02)
Female (base)
Other 2.31 (1.5) 2.15 (1.37)
Male −.71 (.37) −1.21 (.36)**
Age .28 (.02)*** .28 (.02)***
Non-White/White −.40 (.60) −.79 (.57)
High School (base)
Some post-sec. .63 (.70) .31 (.68)
Post-sec. 1.20 (.66) .42 (.65)
Grad school .31 (.72) −.56 (.7)
Other 1.90 (1.2) 1.1 (1.16)
Single (base)
Mar./Common-law 1.06 (.42)* .79 (.4)
Wid./Sep./Div. 1.64 (.77)* 1.34 (.73)
Other .96 (1.2) .89 (1.1)
None (base)
Christian .36 (1.3) −.01 (1.3)
Jewish −4.53 (2.1)* −4.31 (2.1)*
Muslim 1.75 (1.7) 1.73 (1.8)
Other 1.30 (1.4) 1.12 (1.4)
USA (base)
Canada −.32 (.85) −.4 (.81)
UK −4.89 (1.5)** −4.35 (1.5)**
Australia −.99 (1.3) −1.61 (1.3)
Other −.7 (.64) −.2 (.64)
Mom strict .03 (.14) .06 (.12)
Mom warm −.56 (.15)*** −.14 (.14)
Mom talk .24 (.14) −.05 (.12)
Dad strict .19 (.12) .05 (.11)
Dad warm .02 (.15) −.03 (.13)
Dad talk −.26 (.15) −.16 (.13)
Rel. Conflict −1.15 (.08)*** −.83 (.08)***
Rel. Choice −.4 (.10)*** −.04 (.10)
Rel. Importance .31 (.16)* .26 (.15)
Both parents 1.04 (.41)* .14 (.39)
Parents diff. rel −.81 (.49) −1.05 (.46)*
Knew atheists before .74 (.39) 1.99 (.35)***
(Continued)
RELIGION, BRAIN & BEHAVIOR 9
Table 3. Continued.
Block 1 Block 2 Block 3 Block 4 Block 5 Block 6 Block 7
CREDs Only Demo. Variables Parent Quality Religious Variables Relational Variables Institution Variables Complete Model
Peer religiosity 1 −.04 (.04) .12 (.03)***
Peer religiosity 2 .09 (.04)* 0 (.04)
Never stopped att. (base)
Stopped attending −3.50 (.44)*** −2.26 (.41)***
Never attended −8.67 (.69)*** −6.26 (.70)***
Years in rel. school −.08 (.04)* −.04 (.04)
R
2
/ΔR
2
.03/.03 *** .19/.16 *** .03/.01 *** .09/.06 *** .03/.002 * .07/.04 *** .26/.23 ***
Note: ΔR
2
values for each block reflected the difference between the Block 1 (CREDs only) and the current block of interest.
*p< .05, ** p< .01, *** p< .001.
10 J. LANGSTON ET AL.
Because the sample size for the second main analysis was substantially lower than the sample size
for the first analysis, we reran Breusch–Pagan tests, which were again significant for heteroscedas-
ticity, χ
2
= 250.11, p< .001. As a result, we again used robust standard errors (i.e., HC3 corrections)
with coefficients. Within this model, no issues of multicollinearity emerged, with the highest non-
dummy-coded variable having a VIF of 2.51 (M
VIF
= 1.77). When discussing the impact of
CREDs on Age of Atheism, we provided βvalues along with regular B coefficients, as this allowed
for a more intuitive understanding of the CREDs-Age of Atheism relationship.
3. Results
3.1. First set of analyses
Age of Atheism was regressed onto demographic covariates in Block 1 (see Table 2), F(19, 4222) =
28.49, p< .001, R
2
= .159. CREDs was added in Block 2, F(1, 4221) = 85.67, p< .001, R
2
= .177, ΔR
2
= .018, and, as predicted, was associated with a significant delay in the Age of Atheism. Compared to
average CREDs scores, High CREDs scores (+1 SD) were associated with a 1.5-year delay in becom-
ing an atheist. In Block 3, the addition of Religious Importance, Religious Choice, and Religious
Conflict significantly improved the overall model, F(3, 4218) = 76.07, p< .001, R
2
= .218, ΔR
2
= .041. Both Religious Choice and Religious Conflict reduced the predicted Age of Atheism, while
Religious Importance increased it.
Overall, our analyses suggest that CREDs and family religion variables (i.e., Religious Importance,
Religious Choice, and Religious Conflict) predict Age of Atheism. However, the importance of
whether a parent provides consistency in religious messaging and modelling (i.e., CREDs) for pre-
dicting the Age of Atheism, is likely related to other family religion variables. Given these results, we
set out to determine if Religious Importance, Religious Choice, and Religious Conflict moderated the
relationship between CREDs and Age of Atheism. For these analyses, each moderator term was
added after Block 3, meaning that the ΔR
2
for Block 4, Block 5, and Block 6 are in relation to
Block 3.
3.2. Interaction terms
For each interaction term we compared low (−1SD), moderate (M), and high (+1 SD) levels of
Religious Importance, Religious Choice, and Religious Conflict, across low (−1SD), average (M),
and high (+1 SD) levels of CREDs. We report mean differences (M
diff
) in terms of absolute values
for significant interaction terms.
3.2.1. CREDs * religious importance
A moderator term (CREDs*Religious Importance) was added in Block 4, which was statistically sig-
nificant, F(1, 4217) = 13.89, p< .001, R
2
= .221, ΔR
2
= .003, with t= 3.73, p< .001, B= 0.08, 95%
CI [0.04, 0.12]. Thus, the relationship that CREDs had with Age of Atheism changed as a function
of Religious Importance.
At low levels of CREDs (−1SD), we compared Age of Atheism across low (M= 21.84 years),
moderate (M= 22.76 years), and high (M= 23.71 years) levels of Religious Importance. Differences
between these groups were significant, t= 3.56, p< .001, with M
diff
= 0.94 years for high vs. average
levels of Religious Importance, and M
diff
= 1.87 for high vs. low levels for Religious Importance.
At average levels of CREDs (M), we again compared Age of Atheism across low (M= 21.44 years),
moderate (M= 23.19 years), and high (M= 24.95 years) levels of Religious Importance. Differences
between these groups were significant, t= 5.84, p< .001, as well as noticeably larger than at low levels
of CREDs (high vs. average Religious Importance, M
diff
= 1.75 years; high vs. low Religious Impor-
tance, M
diff
= 3.51 years).
RELIGION, BRAIN & BEHAVIOR 11
At high levels of CREDs (+1 SD), we assessed Age of Atheism across low (M= 21.04 years), mod-
erate (M= 23.61 years), and high (M= 26.18 years) levels of Religious Importance. Like previously,
differences were significant, t= 5.64, p< .001, and again larger (high vs. average Religious Impor-
tance, M
diff
= 2.57 years; high vs. low Religious Importance, M
diff
= 5.14 years).
3.2.2. CREDs * religious choice
In Block 5, we added a moderator term for CREDs*Religious Choice, F(1, 4217) = 4.26, p< .001,
R
2
= .219, ΔR
2
= .001, with t=−2.06, p= .039, B=−0.03, 95% CI [−0.06, −0.00]. These results
suggest that the relationship of CREDs to Age of Atheism changes with levels of Religious Choice.
At low levels of CREDs (−1SD), we compared Age of Atheism across low (M= 23.3 years), mod-
erate (M= 23.01 years), and high (M= 22.7 years) levels of Religious Choice. Differences between
these groups were not significant, t=−1.05, p= .293.
At average levels of CREDs (M), we again compared Age of Atheism across low (M= 24.3 years),
moderate (M= 23.6 years), and high (M= 22.98 years) levels of Religious Choice. Differences were
significant, t=−3.39, p< .001 (high vs. average Religious Choice, M
diff
= 0.63 years; high vs. low
Religious Choice, M
diff
= 1.26 years).
At high levels of CREDs (+1 SD), Age of Atheism was assessed across low (M= 25.2 years),
moderate (M= 24.2 years), and high (M= 23.2 years) levels of Religious Choice. As with previous
analyses, differences were significant, t=−4.08, p< .001, and again larger (high vs. average Religious
Choice, M
diff
= 0.99 years; high vs. low Religious Choice, M
diff
= 1.97 years).
3.2.3. CREDs * religious conflict
We removed the moderator term from Block 6, and added the CREDs * Religious Conflict modera-
tor term, F(1, 4217) = 3.91, p= .048, R
2
= .219, ΔR
2
= .001; this moderator term was significant,
t=−1.98, p= .048, B=−.035, 95% CI [−0.06, −0.00]. This suggests that the ability of CREDs to
predict Age of Atheism, varied at low, average, and high level of Religious Conflict.
At low levels of CREDs (−1SD), we compared Age of Atheism across low (M= 24.93 years),
moderate (M= 23.10 years), and high (M= 21.27 years) levels of Religious Conflict. Differences
between these groups were significant, t=−7.89, p< .001, with M
diff
= 1.83 years for high vs. average
levels of Religious Conflict, and M
diff
= 3.67 for high vs. low levels for Religious Conflict.
At average levels of CREDs (M), we again compared Age of Atheism across low (M= 25.92 years),
moderate (M= 23.78 years), and high (M= 21.63 years) levels of Religious Conflict. Differences
between these groups were significant, t=−13.95, p< .001, as well as noticeably larger than at low
levels of CREDs (high vs. average Religious Conflict, M
diff
= 2.15 years; high vs. low Religious Confl-
ict, M
diff
= 4.29 years).
At high levels of CREDs (+1 SD), Age of Atheism was assessed across low (M= 26.91 years), mod-
erate (M= 24.45 years), and high (M= 22.00 years) levels of Religious Conflict. Like previously,
differences were significant, t=−11.88, p< .001, and again larger (high vs. average Religious Conflict,
M
diff
= 2.45 years; high vs. low Religious Conflict, M
diff
= 4.91 years).
Overall, the data analyses revealed a pattern that was not unexpected: while higher levels of
CREDs continued to predict higher Age of Atheism, this was influenced by family religion variables.
In a sense, the “protective”influence of CREDs is stronger in some scenarios (e.g. high Religious
Importance, low Religious Choice, and low Religious Conflict) than in others.
3.3. Second set of analyses
Age of Atheism was regressed onto CREDs in Block 1 (see Table 3), F(1, 3208) = 76.06, p< .001, R
2
= .03, which suggested that CREDs predicted 3% of the variance in Age of Atheism. For CREDs
specifically, t= 8.72, p< .001, B= 0.16, 95% CI [0.12, 0.19], which meant that increasing CREDs
by 1 SD corresponded with a 20-month delay in becoming an atheist. Demographic covariates
were entered in Block 2, F(19, 3189) = 23.18, p< .001, R
2
= .19, ΔR
2
= .16, but CREDs remained
12 J. LANGSTON ET AL.
virtually unchanged, t= 8.79, p< .001, B= 0.15, 95% CI [0.12, 0.19]. Again, increasing CREDs by 1
SD was associated with a 20-month delay in becoming an atheist.
We then removed Block 2 and entered Block 3 (six parental relationship quality items), F(6,
3202) = 4.67, p< .001, R
2
= .03, ΔR
2
= .01, and with this CREDs remained with a similar relationship,
t= 8.70, p< .001, B= 0.17, 95% CI [0.13, 0.21], which was approximately a 22-month delay in the age
of atheism. Generally, demographics and parental relationship quality had limited impact on how
CREDs predicted Age of Atheism.
We then added family predictors in Block 4, F(5, 3203) = 43.83, p< .001, R
2
= .09, ΔR
2
= .06.
CREDs remained significant but experienced its first substantive decline of influence, t= 3.65, p
< .001, B= 0.09, 95% CI [0.04, 0.13]. With the inclusion of family religion predictors [Religious
Conflict, Religious Choice, Religious Importance, Family Structure (lived with both parents), and
Parental Religious Homogamy (whether parents had the same religion or not)], going up 1 SD on
CREDs predicted a delay of just 11 months or so, suggesting that part of CREDs explanatory
power lies in familial structures. We then removed family religion predictors from Block 4 and
entered relational predictors [Peer Religiosity 1, Peer Religiosity 2, and Known Atheist Others] in
Block 5, F(3, 3205) = 2.68, p= .045, R
2
= .03, ΔR
2
= .002, but this did not improve the overall
model. With relational predictors, CREDs remained significant, t= 7.75, p< .001, B= 0.15, 95% CI
[0.11, 0.19], and was associated with a 19-month delay in Atheism.
After removing relational predictors, institutional predictors were entered in Block 6, F(3, 3205)
= 56.20, p< .001, R
2
= .07, ΔR
2
= .04, which were significant. With the inclusion of institutional pre-
dictors [Two-year gap in religious worship service attendance between the ages of 13 and 18, and
number of years of religious schooling during upbringing], CREDs was significant, t= 3.29, p
= .001, B= 0.06, 95% CI [0.02, 0.11], but only predicted a 9-month delay. Given the performance
of institutional predictors, we attended to the pattern of coefficients within this new block. Moving
from No (base) to Stopped attending (before age 13) was especially noteworthy, t=−12.49, p< .001, B
=−8.67, 95% CI [−10.03, −7.31], as it was associated with becoming an atheist 8.5-years earlier.
Despite the obvious conceptual links between ceasing/failing to attend religious worship services
in childhood and the overarching CREDs score, there was only a weak correlation between these
variables (r=−.22), and multicollinearity was decidedly not an issue. Overall, knowing whether a
person had ceased attending religious worship services prior to the age 13 made a far more substan-
tial prediction of Age of Atheism than did CREDs.
As a final step, all variables were entered in Block 7, F(36, 3172) = 24.30, p< .001, R
2
= .26, ΔR
2
= .23, which was significant. With the inclusion of all variables, CREDs was not significant, t= 1.73,
p= .08, B= 0.04, 95% CI [−0.00, 0.09], but given the strong performance of Block 6, this was not
altogether surprising.
4. Discussion
Our first analysis yielded three noteworthy observations. First, Religious Importance predicted a
delay in the Age of Atheism, whereas Religious Choice and Religious Conflict predicted an earlier
onset of Age of Atheism. Second, as hypothesized, CREDs positively predicted Age of Atheism.
Third, there was an interdependency of Religious Importance, Religious Choice, and Religious
Conflict with CREDs.
As Figure 1 shows, as Religious Importance increased within the sample, the age at which a person
became an atheist increased. At low levels of CREDs (−1SD), the difference in the timing of atheism
for persons low or high on Religious Importance was not even two years. In contrast, at high levels of
CREDs, this same difference was over five years. In other words, Religious Importance matters, but
not in isolation: parents espousing the importance of religion may delay atheism in their progeny,
but consistent modelling enhances this effect.
When investigating the relationship between CREDs and Religious Choice, Religious Choice
negatively predicted Age of Atheism. Conceptually, then, the freedom to choose may be one part
RELIGION, BRAIN & BEHAVIOR 13
of a familial atmosphere conducive to atheism earlier in life. However, this interaction term was quite
modest, and, when considering persons at the highest levels of CREDs, the contrast in reported Age
of Atheism between participants reporting low and high levels of Religious Choice was less than two
years’difference (see Figure 2). That said, the pattern of findings was such that higher CREDs were
again associated with a “protective”influence against Age of Atheism, suggesting that even with reli-
gious freedom, high levels of CREDs still “push”respondents toward retaining a belief in god(s).
In the third interaction term, Religious Conflict was negatively associated with Age of Atheism,
while CREDs were again positively associated with Age of Atheism. As can be seen in Figure 3,at
lower levels of CREDs the gap between low and high Religious Conflict was approximately 3.5
years, while at higher levels of CREDs the gap between low and high Religious Conflict was over 5
years. Again, this pattern supports the idea that consistency between messaging and modelling influ-
ences the timing of atheism. In follow-up analysis, we compared low and high CREDs at high levels of
Religious Conflict, and found that there was no difference between Age of Atheism for these two
groups, t=1.51,p= .15. In contrast, when we compared low and high CREDs at low levels of Religious
Conflict, the difference was statistically significant, t= 3.93, p<.001, M
diff
=2.30 years. Framed
0
5
10
15
20
25
30
-1SD M +1SD
Age of Atheism
CREDs
Rel. Import. (-1SD) Rel. Import. (M) Rel. Import. (+1SD)
Figure 1. Differences in age of atheism by CREDs and religious importance.
0
5
10
15
20
25
30
-1SD M +1SD
Age of Atheism
CREDs
Rel. Choice (-1SD) Rel. Choice (M) Rel. Choice (+1SD)
Figure 2. Differences in age of atheism by CREDs and religious choice.
14 J. LANGSTON ET AL.
differently, this pattern of findings suggests that in a high Religious Conflict home, strong religious
modelling is not associated with a delayed Age of Atheism, but in a low Religious Conflict Home it is.
General findings from regression analyses suggested that higher CREDs levels were associated
with delayed Age of Atheism. Functionally, when participants perceived their parents as having
high degrees of religious credibility, this was associated with a “protective”influence against atheism,
and this effect was the most pronounced when Religious Importance was high and Religious Conflict
was low. However, substantive effects were also present at moderate levels of both Religious Impor-
tance and Religious Conflict, but only at high levels of Religious Choice. Also of note, religion of
upbringing had a varied impact on the age at which a person became an atheist. Compared to
those with a nonreligious upbringing, being raised as a Muslim or a religious Other was associated
with a delay in becoming an atheist. However, when controlling for Religious Importance, Religious
Choice, and Religious Conflict, these relationships became non-significant. And yet, even with these
controls, those raised in Judaism were exceptions, and reported an earlier age of atheism (M= 6.02
years) despite the influence of these family-religion variables. Residence in the United Kingdom and
Northern Ireland also displayed a robust and negative association with Age of Atheism, when com-
pared to the much more religious United States. Thus, both religion of upbringing and nationality
should be considered if we wish to understand the formal influence of CREDs and other modern or
current relational dynamics on a person becoming an atheist.
All three family religion variables are consequential for the timing of atheism (although only Reli-
gious Conflict is retained in Block 7 of the second main analysis, thus surviving the influence of all
other [religious] socialization candidate variables), and, among their interrelationships, we can
suggest certain possibilities. To the extent that religion is (perceived as) important in the family,
this may diminish Religious Choice, which in turn may lead to Religious Conflict. However, even
if parents faithfully model their own religion, religion may or may not always be (perceived as)
important in the family of upbringing. We would suggest that whether or not it is, either arrange-
ment is likely to be differentially associated with arrangements of higher/lower Religious Choice
and higher/lower Religious Conflict.
Our second analysis demonstrated that CREDs had a relatively stable ability to predict Age of
Atheism, and remained relevant to the overall model despite the inclusion of demographics, parental
warmth variables, family variables, and relational variables. However, with the inclusion of insti-
tutional predictors, particularly questions about cessation of religious worship service attendance
during one’s adolescent years, CREDs were rendered largely irrelevant to the wider prediction of
Age of Atheism. Yet, we must note that the distribution of CREDs scores across these categories
0
5
10
15
20
25
30
-1SD M +1SD
Age of Atheism
CREDs
Rel. Conflict (-1SD) Rel. Conflict (M) Rel. Conflict (+1SD)
Figure 3. Differences in age of atheism by CREDs and religious conflict.
RELIGION, BRAIN & BEHAVIOR 15
was not uniform: 50% of persons who reported that they had stopped attending religious worship
services prior to the age of 13 scored in the lowest CREDs quintile. When looking at the lowest
two CREDs quintiles this value increased to 75% of all respondents in the category. In other
words, respondents who had stopped attending religious worship services prior to the age of 13,
tended to have parents who scored near the bottom of the overall CREDs measure. We may also
point out that, relative to the Yes and No categories, those in the “Stopped attending prior to age
13”category reported elevated levels of Religious Choice and diminished levels of Religious Conflict.
For these reasons, the performance of the institutional predictors in our analysis does not, in our
view, genuinely overturn the robust contribution of CREDs to Age of Atheism.
4.1. Future Research
Hitzeman and Wastell (2017) suggested that future research should consider both context and con-
tent biases regarding changing levels of religious belief. We second this suggestion, and further
suggest that, if successive generations are becoming less religious (and at younger ages) in post-
industrial countries, this should be examined in reference to changing patterns of (non)religious
socialization (to include levels to CREDs). We have focused primarily on CREDs, but recent work
highlights the role of explicitly nonreligious socialization as an explanation for increasing numbers
of religious “Nones”(which would include atheists) across birth cohorts. Bengtson, Hayward, Zuck-
erman, and Silverstein (2018) argued that some amount of the increasing number of Nones is due to
increasingly explicit nonreligious socialization across generations, that is, not due only to weak or
absent religious socialization. This brings up the question of what kind of CREDs exist in nonreli-
gious families –or if they are needed at all –to sustain whatever positive or substantive worldviews
such parents might wish to instill.
Clearly, family and social environments in which religion is highly salient are not wholly incom-
patible with “religious peace”between family members and peers, or with allowing offspring to make
some Religious Choices for themselves. And yet, this is where a comparative approach might serve
well: Do atheists reflect levels of Religious Choice and Religious Conflict that are different from their
theistic counterparts? Relatedly, were most formerly believing atheists subject to authoritarian, as
opposed to authoritative, parenting styles?
Given weak effects of parental quality items, a more comprehensive measure of parental closeness
and overall family quality might be deployed in future research, as such measures have often been
found to influence the transmission of religious beliefs. Regarding Bengtson et al. (2018) study, it
may be that parental and family quality considerations effectively facilitate both religious beliefs
and atheism across generations. For example, even high-fidelity religious parents may unknowingly
influence the turn to atheism if certain parental qualities and Religious Choice or Religious Conflict
are associated.
4.2. Limitations
First, we only collected data from respondents themselves, and not respondents’parents and peers.
Second, such data inevitably rely upon retrospective self-reports and human memory. This is of par-
ticular concern because as our sample consisted of members of atheist groups, and, just as do “reli-
gious”persons, such individuals use narrative and biography to explain to fellow group members and
others their transition from believer to atheist. And like their religious counterparts, there may be
considerable incentives for atheists to modify, or selectively “re-remember”, their narratives in
order to fit in with their groups. Studying former atheists who had become Christians, Langston
and colleagues (Langston, Albanesi, & Facciani, under review) noted that such conversion narratives
might be just as subject to recall error and retroactive identity construction via biographical narra-
tive, as the narratives of those who stop believing in a god or gods. Effectively, then, in any kind of
(non)religious transition (e.g., from believer to nonbeliever, or vice versa), people are likely to
16 J. LANGSTON ET AL.
“remember”their histories in accordance with creating a positive characterization of their transition,
especially when joining a group or social milieu with preexisting narrative elements which are
learned by converts during transition (Hefner, 1993). Because people are often not fully aware of
the factors that influence their beliefs and behaviours, there is a tendency for rational (or at least con-
scious) reasons to be emphasised, which could probably best be described as “rationalization”of what
are otherwise unknown or unrecognized causal influences on behaviour (Bargh & Chartrand, 1999).
Considered together, such problems underscore the need to very cautiously interpret self-reports and
personal narratives.
Third, we do not know what Religious Conflicts were specifically about in these families, and this
would be important information for determining how, if at all, such conflicts informed later atheism.
A related and critical consideration is whether Religious Conflict is antecedent to or consequent of
atheism. Previous research suggests that one might be as likely as the other (cf. Hunsberger, 1983),
and our design does not allow us to disentangle the temporal order of these two items. While both
orders are logically feasible, our argument rests upon Religious Conflict as an antecedent contri-
bution to becoming an atheist (cf. Pasquale, 2009). Furthermore, even if a person disagrees that
there was family conflict about religion, this does not mean that they do not harbour misgivings
about parental religious beliefs and practices. It is certainly possible that unspoken qualms of
these kinds can manifest in later change and dissent from one’s religious upbringing, especially if
they are suppressed or nursed silently over time. This may be all the more true during the first several
years after a person leaves their parents’home and achieves greater liberation in general, but also
from parental supervision and control (Dudley, 1999).
5. Conclusion
Sociological perspectives are often overlooked in the cognitive science of religion. Yet, families, social
relationships, and (religious) institutional contexts have long been understood as primary sites of
religious socialization processes which are influential in religious outcomes. As such, work in CSR
should be integrated with such perspectives. Because discussing changes in the modern religious
landscape also requires a consideration of where atheism is increasing (Norris & Inglehart, 2011),
it is important to note broad social and intergenerational change in such societies, especially as
they impact (and are impacted by) broader changes in family formation patterns and socialization
processes. This also fits well regarding considerations of context and content biases, and with con-
siderations of both ultimate and proximate mechanisms that underpin (non)religious phenomena.
Our study provides a glimpse into how CREDs work in tandem with other religion variables func-
tioning in family and relational processes, and we envision it as a way forward in linking the cogni-
tive science of religion to more sociologically oriented approaches to (non)religious outcomes.
Notes
1. In Pasquale’s study of secular group affiliates, of those current secularists who were raised in a Roman Catholic
background (n= 128), 37.5% reported conflict with parents over religion. For those with a Protestant back-
ground (n= 469), 24.3% reported such conflict, whereas for those raised with a secular background (n=
105), this figure was only 2.9%.
2. A majority of these studies specifically measure parental CREDs, and not CREDs in general (e.g., religious auth-
ority figures, relatives, members of the community). See also Turpin, Andersen, & Lanman, 2018; while they
found no immediate effects of CRED exposure on measures of explicit and implicit belief using an experimental
manipulation, they did find further correlational evidence linking past CRED exposure to self-reported reli-
gious belief.
Disclosure statement
No potential conflict of interest was reported by the authors.
RELIGION, BRAIN & BEHAVIOR 17
ORCID
Joseph Langston http://orcid.org/0000-0002-6026-0272
David Speed http://orcid.org/0000-0001-7033-2068
Thomas J. Coleman III http://orcid.org/0000-0002-3003-5090
References
Bandura, A. (1971). Social learning theory. Morristown, NJ: General Learning Press.
Bandura, A. (2003). Commentary: On the psychosocial impact and mechanisms of spiritual modeling. International
Journal for the Psychology of Religion,13(3), 167–173.
Bargh, J. A., & Chartrand, T. L. (1999). The unbearable automaticity of being. American Psychologist,54(7), 462–479.
Barrett, J. L. (2007). Cognitive science of religion: What is it and why is it? Religion Compass,1(6), 768–786.
Barry, C. M., & Christofferson, J. L. (2014). The role of peer relationships in emerging adults’religiousness and spiri-
tuality. In C. M. Barry & M. M. Abo-Zena (Eds.), Emerging adults’religiousness and spirituality: Meaning-making in
an age of transition (pp. 76–92). New York, NY: OUP.
Bengtson, V. L., Hayward, R. D., Zuckerman, P., & Silverstein, M. (2018). Bringing up nones: Intergenerational influ-
ences and cohort trends. Journal for the Scientific Study of Religion,1–18. Retrieved from https://onlinelibrary.wiley.
com/doi/abs/10.1111/jssr.12511.
Bengtson, V. L., Putney, N., & Harris, S. (2013). Families and faith: How religion is passed down across generations.
New York, NY: OUP.
Berkers, E. (2018, in press). The atheistic factor? Explaining the link between being atheistic and childrearing values in
30 countries in Europe. European Societies.
Bruce, S., & Glendinning, T. (2010). When was secularization? Dating the decline of the British churches and locating
its cause. The British Journal of Sociology,61(1), 107–126.
Clark, C. A., & Worthington Jr., E. L. (1990). Family variables affecting the transmission of religious values from
parents to adolescents: A review. In B. K. Barber & B. C. Rollins (Eds.), Parent-adolescent relationships (pp.
167–191). Lanham, MD: University Press of America.
Cliteur, P. B. (2009). The definition of atheism. Journal of Religion and Society,11,1–23.
Dudley, R. L. (1999). Youth religious commitment over time: A longitudinal study of retention. Review of Religious
Research,41(1), 110–121.
Flor, D. L., & Knapp, N. F. (2001). Transmission and transaction: Predicting adolescents’internalization of parental
religious values. Journal of Family Psychology,15(4), 627–645.
Garcia, A., & Blankholm, J. (2016). The social context of organized nonbelief: County-level predictors of nonbeliever
organizations in the United States. Journal for the Scientific Study of Religion,55(1), 70–90.
Gervais, W. M., & Henrich, J. (2010). The zeus problem: Why representational content biases cannot explain faith in
gods. Journal of Cognition and Culture,10(3), 383–389.
Gervais, W. M., & Najle, M. B. (2015). Learned faith: The influences of evolved cultural learning mechanisms on belief
in gods. Psychology of Religion and Spirituality,7(4), 327–335.
Gervais, W. M., Willard, A. K., Norenzayan, A., & Henrich, J. (2011). The cultural transmission of faith: Why innate
intuitions are necessary, but insufficient, to explain religious belief. Religion,41(3), 389–410.
Hefner, R. W. (Ed.). (1993). Conversion to christianity: Historical and anthropological perspectives on a great transform-
ation. Berkeley, CA: University of California Press.
Henrich, J. (2009). The evolution of costly displays, cooperation and religion. Evolution and Human Behavior,30(4),
244–260.
Hitzeman, C., & Wastell, C. (2017). Are atheists implicit theists? Journal of Cognition and Culture,17(1-2), 27–50.
Hunsberger, B. E. (1983). Apostasy: A social learning perspective. Review of Religious Research,25,21–38.
Hunsberger, B., Pratt, M., & Pancer, S. M. (2002). A longitudinal study of religious doubts in high school and beyond:
Relationships, stability, and searching for answers. Journal for the Scientific Study of Religion,41(2), 255–266.
King, P. E., Furrow, J. L., & Roth, N. (2002). The influence of families and peers on adolescent religiousness. Journal of
Psychology and Christianity,21(2), 109–120.
Langston, J., Albanesi, H., & Facciani, M. (under review). Toward faith: A qualitative study of how atheists convert to
christianity.
Lanman, J. A. (2010). A secular mind: Towards a cognitive anthropology of Atheism (Doctoral dissertation). University
of Oxford.
Lanman, J. A. (2012). The importance of religious displays for belief acquisition and secularization. Journal of
Contemporary Religion,27(1), 49–65.
Lanman, J. A., & Buhrmester, M. D. (2016). Religious actions speak louder than words: Exposure to credibility-enhan-
cing displays predicts theism. Religion, Brain & Behavior,7(1), 3–16.
Mahoney, A. (2005). Religion and conflict in marital and parent-child relationships. Journal of Social Issues,61(4), 689–706.
18 J. LANGSTON ET AL.
Maij, D. L., van Harreveld, F., Gervais, W., Schrag, Y., Mohr, C., & van Elk, M. (2017). Mentalizing skills do not differ-
entiate believers from non-believers, but credibility enhancing displays do. PloS one,12(8), e0182764.
Manning, C. (2015). Losing our religion: How unaffiliated parents are raising their children. New York, NY: NYU Press.
Norris, P., & Inglehart, R. (2011). Sacred and secular: Religion and politics worldwide. New York, NY: Cambridge
University Press.
Okagaki, L., Hammond, K. A., & Seamon, L. (1999). Socialization of religious beliefs. Journal of Applied Developmental
Psychology,20(2), 273–294.
Pasquale, F. L. (2009). A portrait of secular group affiliates. In P. Zuckerman (Ed.), Atheism and secularity (Vol. 2, pp.
43–88). Denver, CO: Praeger.
Petts, R. J. (2014). Parental religiosity and youth religiosity: Variations by family structure. Sociology of Religion,76(1),
95–120.
Potvin, R. H., & Sloane, D. M. (1985). Parental control, age, and religious practice. Review of Religious Research,27(1),
3–14.
Putney, N. M., Lam, J. Y., Nedjat-Haiem, F., Ninh, T. H., Oyama, P. S., & Harris, S. C. (2013). The transmission of
religion across generations: How ethnicity matters. In M. Silverstein & R. Giarrusso (Eds.), Kinship and cohort
in an aging society: From generation to generation (pp. 209–236). Baltimore: Johns Hopkins.
Smith, C., & Denton, M. L. (2005). Soul searching: The religious and spiritual lives of American teenagers. New York,
NY: Oxford University Press.
Thiessen, J. (2016). Kids, you make the choice: Religious and secular socialization among marginal affiliates and non-
religious individuals. Secularism and Nonreligion,5(1), 1–16.
Thiessen, J., & Wilkins-Laflamme, S. (2017). Becoming a religious none: Irreligious socialization and disaffiliation.
Journal for the Scientific Study of Religion,56(1), 64–82.
Tratner, A. E., Sela, Y., Lopes, G. S., Ehrke, A. D., Weekes-Shackelford, V. A., & Shackelford, T. K. (2017). Individual
differences in childhood religious experiences with peers. Personality and Individual Differences,119,73–77.
Turpin, H., Andersen, M., & Lanman, J. A. (2018). CREDs, CRUDs, and Catholic scandals: Experimentally examining
the effects of religious paragon behavior on co-religionist belief. Religion, Brain & Behavior,87,1–13. doi:10.1080/
2153599X.2018.1439087
Voas, D., & Doebler, S. (2011). Secularization in Europe: Religious change between and within birth cohorts. Religion
and Society in Central and Eastern Europe,4(1), 39–62.
Voas, D., & Doebler, S. (2014). Secularization in Europe: An analysis of inter-generational religious change. In W. Arts
& L. Halman (Eds.), Value contrasts and consensus in present-day Europe: Painting Europe’s moral landscapes (pp.
231–248). Leiden: Brill.
West, S. G., Aiken, L. S., & Krull, J. L. (1996). Experimental personality designs: Analyzing categorical by continuous
variable interactions. Journal of Personality,64(1), 1–48.
Willard, A. K., & Cingl, L. (2017). Testing theories of secularization and religious belief in the Czech Republic and
Slovakia. Evolution and Human Behavior,38(5), 604–615.
Xygalatas, D. (2014). Cognitive science of religion. In Encyclopedia of psychology and religion (pp. 343–347). New York,
NY: Springer.
Zuckerman, P. (2012). Contrasting irreligious orientation: Atheism and secularity in the USA and Scandinavia.
Approaching Religion,2(1), 8–20.
Appendix
CREDs Exposure Scale (see Lanman & Buhrmester, 2016)
Instructions: The following questions ask about experiences during your upbringing that relate to religion. Specifically,
the questions ask about your perceptions of your primary caregiver or caregivers (i.e., parents or guardians). Please
answer each of the following according to your overall impression of your caregiver(s) on the following scale:
1234567
To no extent at all To an extreme extent
(1) To what extent did your caregiver(s) attend religious services or meetings?
(2) To what extent did your caregiver(s) engage in religious volunteer or charity work?
(3) Overall, to what extent did your caregiver(s) act as good religious role models?
(4) Overall, to what extent did your caregiver(s) make personal sacrifices to religion?
(5) To what extent did your caregiver(s) act fairly to others because their religion taught them so?
(6) To what extent did your caregiver(s) live a religiously pure life?
(7) To what extent did your caregiver(s) avoid harming others because their religion taught them so?
RELIGION, BRAIN & BEHAVIOR 19
