2.1 Measuring Religion

Since religion is a complex thing, scholars find it helpful to distinguish between different dimensions of personal religiosity. Various ways of grouping these have been proposed. One helpful framing is Saroglou’s ‘four Bs’ scheme, which includes belonging, believing, behaving, and bonding. ‘Belonging’ primarily refers to being a member or follower of a given religious group, including whether and how strongly a person identifies or affiliates in religious terms. ‘Believing’ most obviously encompasses specifically religious doctrines (existence of God, karma, etc.), but also religion-relevant moral or social attitudes (abortion, the permissibility of eating meat). ‘Behaving’ and ‘Bonding’ both refer to behaviour. ‘Bonding’ concerns directly religious activities, such as attending religious services, praying, or reading certain scriptures. The use of the term implies that such activities are intended to help practitioners connect with a transcendent reality, other co-religionists, and/or the self. ‘Behaving’ includes other sorts of activities, like volunteering, giving to charity, or drinking alcohol, that might be religiously encouraged or discouraged.

However one chooses to cut and label the religious cloth, these different dimensions do not always match up perfectly at the level of individuals. That is to say, a person who belongs to a given religious group (e.g., identifies as a Roman Catholic) might not necessarily believe (e.g., in transubstantiation), behave (e.g., give alms to the needy), and/or bond (e.g., through attending Mass and receiving Holy Communion) in predetermined ways that are normative to that religious group. Likewise, not having a religious affiliation (i.e., identifying as, in some way, nonreligious) need not stop a person from, say, holding Hindu-inspired beliefs about karma and reincarnation, or attending services at the local synagogue. Nevertheless, in the aggregate, these different aspects of religiosity are far from independent, even if their co-occurrence varies widely depending on the context (Saroglou et al., Reference Saroglou, Clobert, Cohen, Johnson, Ladd, Van Pachterbeke and Tapia Valladares2020). The relationships among these aspects can also sometimes be asymmetrical. For example, knowing that a person attends church regularly, or believes strongly in the Virgin Birth of Jesus, is normally a strong predictor that they identify as some kind of Christian. But does it work in the other direction? In many Western countries, not so much: large numbers of Christian ‘belongers’ neither believe nor bond in strongly, distinctively Christian ways. In somewhere like China, much more so: identifying as a Christian is likely a very strong indicator that they believe and practise (if and when able to).

Given that ‘religion is a multidimensional variable composed of several facets’ (Paloutzian, Reference Paloutzian2017, p. 26), this means that when trying to measure it, one has to decide not only which dimensions, but which specific aspects of those dimensions, to focus on. In practice, this requires making judgements about which one or more elements are the most important or salient or telling. Hence, studies of intergenerational transmission must focus on a discrete set of things being passed on (or not) over a period of time. Some studies focus on belonging alone. This is especially true in sociology. For example, the rise of the ‘nones’ in the USA is primarily about changing religious belonging – driven largely by people raised religiously by religious parents who, in adult life, come to answer ‘no religion’ on surveys (Burge, Reference 68Burge2021; Bullivant, Reference Bullivant2022).

Of course, scholars trying to understand this phenomenon will be interested in all sorts of other aspects of religiosity too. But note that the analytic focus here is squarely on the failure to successfully transmit a religious affiliation from one generation to the next. However, as noted earlier, belonging alone does not always tell you much about other aspects of a person’s religiosity. For this reason, studies often combine belonging with a bonding or believing measure. That is, they are trying to ascertain whether, and to what extent, a person is a ‘believing X’ or a ‘practising X’, rather than just an (self-identifying) ‘X’. In truth, this kind of approach captures much more of what religious groups, and indeed many religious parents, are typically aiming-at or hoping-for in raising children. Of course, different religious traditions will have their own implicit shorthands for what successful transmission looks like: an observant Jew, a Mass-going Catholic, a Bible-believing Evangelical. Studies focused on specific groups will likely, therefore, be tailored to fit what few aspects of religiosity are deemed most meaningful, or revealing, about a person’s overall religiosity in a given setting.

Studies designed for one religious setting might not, however, translate very easily to others. Most obviously, while attendance at religious services is a crucial measure of religious behaviourFootnote ³ in Christian, Jewish, and Muslim contexts (although again, precise expectations differ both across and within these broad traditions), it means much less in many Buddhist, Hindu, or Daoist ones. More wide-ranging studies, interested in religious transmission in general and/or across a range of contexts, typically need to be more explicitly multidimensional. For example, Van Tongeren and DeWall (Reference Van Tongeren and DeWall2023) have proposed a framework for exploring the loss of religion in multidimensional terms, explicitly patterned on Saroglou’s four Bs, rather than simply as a matter of no longer religiously identifying (i.e., becoming a ‘none’) or believing (e.g., in the existence of God). Generally speaking, however, these are exceptions. For reasons both of resource (time, money, energy) and analytic focus/neatness, a large proportion of the transmission literature is either mono- or bidimensional. There is nothing illegitimate about this, and indeed the general co-occurrence of dimensions often means that measuring one aspect of religiosity is usually a decent proxy for others. But it is still worth asking, when reading a study on the transmission of religion: which bit or bits of religion, exactly, are actually being measured here? In this Element, we will try to be specific about this, but will not always manage to do so for the sake of concision.

More briefly, it is also always worth asking how it is being measured. Asking good questions about religion is not always easy, and there is a large literature devoted to the topic (Hackett, Reference Hackett2014; Hill et al., Reference 73Hill, Hood, Lewis-Jong and Harrisforthcoming). Even seemingly simple matters, such as probing a respondent’s religious affiliation, can be more complicated than they first seem: different wordings can nudge a surprising number of people to answer apparently equivalent questions in different ways (Field, Reference Field2014). For various reasons, survey respondents are not always reliable reporters of their own present religious practice (Hadaway et al., Reference Hadaway, Marler and Chaves1998; Brenner, Reference Brenner2014). That fact might give us pause for thought regarding their reliability in reporting either their own, or their parents’, levels of religiosity decades previously. This is especially so if, as is likely, such memory and reporting biases are influenced by present-day religiosity. Note too that, in cross-national studies, key terms – not least ‘god’ and ‘religion’ – do not necessarily translate in straightforward ways (Tanaka, Reference Tanaka2010; Höllinger & Eder, Reference Höllinger and Eder2016). These issues are endemic to the social science of religion and therefore come baked into transmission research. Their ubiquity does not, however, absolve either researchers themselves, or those reading and interpreting their productions, from being mindful of them.

2.2 Study Design: Cross-Sectional Surveys

The very idea of intergenerational transmission implies the passage of time. At the very least, this requires measurement of (though not necessarily at) a Time A and Time B: think ‘before’ and ‘after’, ‘past’ and ‘present’, ‘how it started’ versus ‘how it’s going’. In the study of the intergenerational transmission of religion, what is measured at Time A ought, strictly speaking, to be parental religiosity – that is, the thing to be passed on. Whereas at Time B, we are instead interested in the offspring’s religiosity, usually at some point in their adult life – this is to assess whether, in what ways, and to what extent, the parents’ religiosity has been passed on to the children, when the latter are no longer entirely dependent on the former. This is the basic blueprint for all social-scientific research into intergenerational religious transmission, albeit one that allows for a good deal of adaptation, elaboration, and innovation. This will hopefully be evident from the studies we cite and discuss throughout this Element. One major and important variant is worth highlighting here, though. Since transmission is a dynamic process, it is often desirable to measure the child’s religiosity at a number of intervals (Times A₁, A₂, A₃…), such as in early childhood, adolescence, and emerging adulthood. These measures sit between the ‘before’ and the ‘after’, and can be treated as one or the other, depending on the context of the study. Sometimes, researchers are interested in whether people, as adults, have retained the religiosity of their upbringing, which very often reflects the religiosity of their parents.Footnote ⁴

In common with much psychological research, most study designs can usefully be divided into one of two types: cross-sectional and longitudinal. The former is, by far, the most common. This may seem surprising, given what we said earlier about the requirement for (at least) two separate time points. After all, ‘the collection of data at a single point in time’ (Krosnick et al., Reference Krosnick, Lavrakas, Kim, Reis and Judd2014, p. 406, emphasis added) is a large part of the textbook definition of a cross-sectional study. Thus, researchers typically ask participants questions about both their present religiosity (i.e., Time B), and their and/or their parents’ past religiosity (i.e., Time A) at some point during their childhood, relying in the latter case on their memory of the past. Furthermore, although most cross-sectional studies are ‘one-off’ surveys, some are repeated with different but comparable (e.g., demographically representative) groups of individuals over time: these studies do not tell us how individuals or families change over time, but may tell us how societies, or cohorts within those societies (see Section 3.3), change over time.

The International Social Survey Programme (ISSP) is a good example of a project that annually collects cross-sectional data and which sometimes includes questions relevant to the intergenerational transmission of religion. The ISSP does not solely focus on religion, but has – through its alliance of national surveys, such as the General Social Survey in the USA and the British Social Attitudes survey in the UK – periodically administered a dedicated ‘Religion’ module. Thus far, it has done so four times – in 1991, 1998, 2008, and 2018 – with a fifth wave currently planned in 2028. Since 1984, the ISSP has included fifty-seven countries, but the precise number of countries represented in each wave fluctuates. For example, the 2008 and 2018 modules were fielded in an extra fourteen countries, mostly in African and Asian countries, where the ISSP (and the social-scientific study of religion more generally) is under-represented.

The World Values Survey (WVS; including the European Values Survey) started in 1981 and has covered over 100 countries across its seven completed waves (with Wave 8 currently underway), though like the ISSP, the number of countries represented in each wave varies. Like the ISSP, it is a repeated cross-national, cross-sectional survey, but it also differs from the ISSP in significant ways. Rather than running annually, it is deployed in fewer waves that each take a longer time (e.g., Wave 7 was collected between 2017 and 2022). Also unlike the ISSP, although the questionnaire does change somewhat between waves, the waves are not themed. This is in part because the overall focus of the WVS is narrower, targeting beliefs and values; on the other hand, the ISSP tends to ask more detailed questions within each theme. Consequently, there have been religiosity items in every wave of the WVS, but these are less detailed than in the ISSP.

Both the ISSP and WVS have included questions about present and ‘raised’ religion. Besides asking about current religiosity across a number of dimensions (e.g., religious affiliation, religious service attendance, beliefs in God, and life after death), the WVS also asks whether the respondent was raised religiously and how important they think it is for children to learn religious faith at home. As noted above, the ISSP’s Religion module asks a much wider set of questions. These include questions about a variety of beliefs and attitudes, diverse types of religious practice, and – critically here – the religious affiliation and attendance of each parent separately (when the respondent was a child), and how often the respondent themselves attended religious services when they were around 11 or 12 years old.

These questions hardly capture the fullness of a person’s religious upbringing, but there is still much that can be profitably done with the resulting data. Certainly, with WVS and ISSP data, we can look at how religion has changed at a societal level over time, including how different religious/denominational affiliations have changed, and religious beliefs and behaviours of different kinds too. Questions about whether societies are secularising – and in what way – are answerable in part by relying on these kinds of data. At an individual level, we can also ascertain the retention levels of different religious groups by comparing ‘raised’ and present religiosity: X% of people raised mainline Protestant now say they have no religion, Y% of cradle Catholics attend church at least monthly, Z% of evangelical Protestants are converts (i.e., were raised as something else). Given that these surveys are repeated and collect data from adults of different ages, it is also possible to track these retention rates over time – how Boomers differ from Gen Zs, and so on – at least to some extent. Sociologists, especially, love doing this kind of thing (e.g., Sherkat, Reference 81Sherkat2014; Bullivant, Reference Bullivant2019), and the basic idea can be profitably extended to explore religious change cross-culturally (e.g., Lesage et al., Reference Lesage, Starr and Miner2025, based primarily on Pew Center data comparing ‘raised’ versus ‘current’ religious affiliation in over thirty countries). This basic method can answer a variety of questions. For example, and as we shall cover in more detail later, researchers have relied on these data to examine the effect of national-level religiosity/secularity on intergenerational transmission (see Section 4.1) and to examine the influence of religious differences between two parents on their child’s overall religiosity (see Section 5.5).

Naturally, even the ISSP’s Religion module only goes so far in terms of detail. One survey (or in this case, one bolt-on battery of questions to another survey) can only cover so much ground. It also needs to ask questions that are sufficiently generic to apply very widely across religious traditions and national contexts. Even the largest national surveys, moreover, yield small numbers of specific subgroups that might be of interest. As important as such large-scale ‘big trend’ type research is, it rarely gets us beyond very general insights. More finely grained research, therefore, relies heavily on cross-sectional surveys targeted at much narrower groups: just young adults, or only Muslims, say. This allows for much richer, detailed, and context-sensitive questions about both past and present religiosity to be asked and answered, which is clearly a good thing. Religious transmission never happens ‘generically’: it is always specific forms of religious belonging, believing, bonding, and behaving that are transmitted (or not) in specific ways. So if we want really to understand, say, Jewish transmission, then we must ask specifically Jewish questions to a sufficient number of Jews and/or people who were raised Jewish (James et al., Reference James, Lester and Brooks2012; Rubin & Rubin, Reference 80Rubin and Rubin2014). (We can go even further too: if we really want to understand Hasidic Jewish transmission in Israel…) Of course, such specificity may sacrifice easy extrapability: what works in transmitting one form of religion in a particular time and place, may not work in transmitting a different form of religion in that same time and place, or the same form of religion in a different one. For this reason, in this Element we rely mainly on more general studies, but also try to cover some more specific cases (see especially Section 6).

2.3 Study Design: Longitudinal Studies

Cross-sectional studies with retrospective questions about the respondent’s own, and/or their parents’, religiosity during upbringing are a convenient shortcut to gaining data on Time A. But it would be far better to collect data at Time A itself, and then to conduct a follow-up study, years later, with the same participants at Time B. Even better would be to collect data at several points during childhood and even into ‘emerging adulthood’ (18–25 years; Arnett, Reference Arnett2000) in order to more fully chart both changing religiosity over these years (a significant topic in its own right) and capture a much broader range of possible influences on fully-fledged adult religious outcomes. And why stop there? After all, religiosity is not static during adulthood (see Section 3.1). In an ideal world, we would keep on following-up with the original participants at regular intervals ‘till death do us part’.

This is the essential logic of longitudinal studies of religious transmission. They allow us to track religiosity in something approaching real time, without having to trust respondents’ memories of ‘when you were eleven or twelve’ or suchlike, and to do so in the context of social, cognitive, and personality factors. Childhood and adolescent religious development does not happen within a vacuum (see Lerner & Lerner, Reference Lerner, Lerner, McAdams, Shiner and Tackett2019; Mahoney, Reference Mahoney2021). Quite apart from the greater immediacy involved in studying a person’s upbringing during that upbringing, separate waves of data collection typically allow more time to ask a fuller set of questions. A further benefit flows from the fact that research with children necessarily requires the involvement of their parents/caregivers. At a minimum, they have to give their formal consent. For research at the younger age range, they may also be needed to answer questions either for, or with, their children. This opens up the possibility of accessing parental religiosity at the source, so to speak: surely preferable to relying on the retrospection of their grown-up children. As we shall see, this golden opportunity is one that some researchers have seized with both hands.

For all its positives, longitudinal research involves very significant practical challenges. While all good social science requires a significant outlay of time, effort, and (often) money, longitudinal research requires committing all three for years, if not decades, in advance. Funding must be secured for a minimum of two waves, often at a space of several years. Even if this very high bar is met, funding for hoped-for additional waves is likely to remain ‘TBC’ for a long time, and perhaps indefinitely. It also requires much more from participants. Keeping track of them over long periods is a major challenge in itself. Persuading them to keep re-engaging is critical. A high rate of attrition not only reduces the overall n (sample size) of the study but also generates concerns of ‘survivorship bias’ in the remaining sample: that is, do those who drop out differ in systematic ways from those who do not, and how might this skew the resulting data (cf. Hayward & Krause, Reference Hayward and Krause2016)? A high degree of buy-in from participants also has its upside, however. Having participated once, they may well be more willing to re-participate a second (or third, or fourth) time than they would if ‘cold-called’; a nice example of the foot-in-the-door effect (Freedman & Fraser, Reference Freedman and Fraser1966) at work.

If one does not simply embark on longitudinal research, then we can at least be grateful that some intrepid souls do indeed do so – and more to the point, did so several decades ago. For in the words of Vern L. Bengtson and colleagues:

Bengtson should know. His own Longitudinal Study of Generations (LSOG) began life in 1970 as a cross-sectional study of three generations – grandparents, parents, and (adolescent/young-adult) children – from the same families, living in southern California. Then, with funding to do a follow-up study fourteen years later, a longitudinal research programme was born. Waves 2 through 7 came at three-yearly waves between 1985 and 2000, Wave 8 was collected in 2005, Wave 9 in 2016, and Wave 10 in 2022. A fourth generation, comprising the great-grandchildren of the original grandparents, was added in 1991, after these individuals turned 16 (Bengtson et al., Reference Bengtson, Putney and Harris2013; Brown et al., Reference Brown, Blankholm, Hoesly, Hwang, Harris, Silverstein, Silverstein, Gärtner and Brown2024).Footnote ⁵

Given both the relative rarity of longitudinal research and its significance to the field, it is worth mentioning two other important studies here. Findings from both, in common with LSOG, will be cited and discussed throughout this Element. The National Study of Youth and Religion (NSYR) first surveyed a nationally representative sample of American 13 to 17 year olds (plus an oversample of Jewish teens), along with their parents, in 2003. Follow-up waves with the original youths (but not their parents) were conducted in 2005, 2007/2008, and 2013/2014, when participants were aged 23 to 28 years, thus capturing the cohort’s full span of adolescence and emerging adulthood.

While both LSOG and NSYR were created with religious transmission in mind, the Avon Longitudinal Study of Parents and Children (ALSPAC) was instead ‘specifically designed to determine ways in which the individual’s genotype combines with environmental pressures to influence health and development’ (Iles-Caven et al., Reference Iles-Caven, Bickerstaffe, Northstone and Golding2024, p. 3). Centred on Bristol, in south-west England, its founding cohort consisted of over 14,000 expectant mothers. These, their offspring, and the mothers’ partners have been surveyed at various intervals ever since. Although not the focus, some religious data was collected for both parents and children in each wave (including at multiple points during the children’s childhoods) as part of standard ‘background info’. Building on this crucial bedrock of longitudinal data, a dedicated survey on ‘Religion and Spiritual Belief and Behaviour’ was then administered to both parents and offspring (now aged 28 to 29) in 2019/2020 (Iles-Caven et al., Reference Iles-Caven, Bickerstaffe, Northstone and Golding2024; Major-Smith et al., Reference Major-Smith, Morgan, Halstead, Tohidinik, Iles-Caven, Golding and Northstone2023).

2.4 Qualitative and Mixed-Methods Approaches

As noted earlier, in many studies of religious transmission (some of our own not excluded…), a single tick-box question along the lines of ‘In what religion, if any, were you raised?’ has to do an awful lot of heavy lifting. The load can be shared, at least to some extent, by a fuller arsenal of questions: even knowing the extent to which a person’s upbringing as an ‘X’ involved, say, attending religious services, tells us a lot more. Even so, the full richness and complexity of a person’s religious upbringing are hardly captured by the sorts of measures typically used by psychologists. This is especially true for studies working cross-traditions: the very real differences between being raised, say, a practising Evangelical, or a practising Orthodox Jew, or a practising Jehovah’s Witness, are apt to be elided or ignored by purely quantitative measures. This observation is not unique to either transmission research, or the psychology of religion in general. The past several decades have seen a significant growth in qualitative research across many areas of psychology, with various types of interview-based methods most prominent (Coyle, Reference Coyle, Lyons and Coyle2021). Such approaches have been perfectly respectable for much longer within sociology. Between the two disciplines, then, qualitative research has significantly enriched our understanding of intergenerational transmission.

Of course, despite how the distinction between quantitative and qualitative work is often framed in methodology textbooks (i.e., as implying some fundamental difference of epistemological ‘paradigm’), in practice a good deal of work is both. For example, it is perfectly common for researchers to unite both analysis of (‘quant’) survey data, with (‘qual’) interviews or focus groups, and perhaps even (also ‘qual’) participant observation and ethnography (techniques more closely associated with anthropological research). Hence, Bullivant’s (Reference Bullivant2022) study of American ‘nonverts’ – i.e.that is, people raised as some kind of Christian, but who now identify as ‘nones’ – uses data from the General Social Survey to provide a statistical skeleton of the phenomenon as a whole, which is then significantly fleshed out with interview data. Riess and Knoll’s (forthcoming) book on leaving Mormonism draws on the authors’ own nationally representative surveys of US cradle Latter-day Saints, plus around a hundred interviews. Qualitative research can also be profitably integrated into longitudinal studies. Notably, the afore-mentioned National Study on Youth and Religion incorporated in-depth interviews of a subsample of those surveyed in each of its four waves, yielding multiple significant mixed-method publications (e.g., Smith et al., Reference Smith, Longest, Hill and Christofferson2014; Denton & Flory, Reference Denton and Flory2020).

Qualitative research need not be paired directly with quantitative research in order to be valuable, even if there is often a more indirect quantitative informing of such work. For example, researchers might choose to focus on a particular group because previous survey-based studies have shown there is something especially interesting about them, transmission-wise (e.g., they have better or worse outcomes), or ask questions on particular topics based on what the statistical analyses reveal.

Qualitative approaches can also profitably be deployed to ‘refresh the parts other methods don’t reach’ (to paraphrase an old beer commercial). For instance, given the paramount importance of parents and parenting to transmission (see Section 5), it is noteworthy that very few quantitative studies involve them directly, instead usually relying on their grown-up children’s remembered testimonies about them. (There are exceptions, as we have already noted.) As such, in-depth interviews of parents while they are actually parenting have a great deal of value. There have recently been important studies of how both religious (e.g., Smith et al., Reference Smith, Ritz and Rotolo2019) and nonreligious parents (Manning, Reference Manning2015) actively seek to pass on their own religiosity or nonreligiosity to their offspring. Likewise, talking to pre-teen children about religion while they are still pre-teens (Strhan & Shillitoe, Reference Strhan and Shillitoe2025), and/or observing them in key sites of socialisation such as school religious assemblies (Shillitoe, Reference Shillitoe2023), is also undoubtedly appealing.

Another important topic which large surveys typically miss is the members of small religious groups and/or new religious movements (NRMs). While there have been survey-based studies of transmission within certain groups (e.g., Inose, Reference Inose2005, on the Buddhist NRM, Sōka Gakkai, in Japan), the particularity of a religious upbringing within such groups is especially amenable to qualitative exploration. Most notable here is van Eck Duymaer van Twist’s (Reference Van Eck Duymaer Van Twist2015) book-length exploration, based on interviews with both parents and (now grown-up) children raised in several NRMs: the Church of Scientology, the Bruderhof, International Society for Krishna Consciousness (Hare Krishnas), the Unification Church, and the Family International. While very different, these groups all exhibit high levels of intensive and/or segregated socialisation, highly committed parents and peers, and a high degree of difference from the religious, social, and (often) moral norms of their host societies: all factors significant to transmission outcomes. Hence, while interesting and worthy of study in their own right, they also arguably have a wider significance for theory generation as comparatively ‘extreme’ test cases.

This Element will mainly rely on quantitative research, but some of the projects from which we draw do include qualitative methods (e.g., NSYR); and where it is necessary or valuable, we do also discuss smaller scale interview-based research.

3.1 Ageing and the Life Course

Religiosity does not normally remain static over the course of an adult lifetime: ‘to live is to change’ (Newman Reference Newman1845, p. 39), to quote someone whose own biography is a good example. People’s beliefs, and/or the strength or confidence with which they hold them, change over time. (The same is also true of moral, social, and political beliefs, which are often also themselves religious or religion-adjacent.) Habits of worship, prayer, and/or devotional activities fluctuate. More rarely, people change their religious affiliations and identities, whether formally (i.e., officially joining or leaving a given religion, denomination, or group) or informally (e.g., feeling that a different religious label – none, Buddhist, pagan – is a now truer fit, and ticking that box on surveys instead).

Why people become religious, increase or decrease their existing religiosity, change religion, or abandon it, are all interesting areas of social-scientific research. Here, though, we are interested in the aggregate effects of all this. Do people, on average, change their religiosity in predictable ways over their lifetimes? Or, to put it more simply, do people generally become more or less religious as they age? Methodologically, there are two main ways of finding out. Longitudinal research is an obvious help, though since most long-term studies do not use nationally representative samples, generalisability is an issue here. Repeated cross-sectional surveys, on the other hand, if done carefully, allow researchers to track particular cohorts over time (albeit with different respondents in each wave).

Based on a large number of studies, there is a consensus that people do generally get more religious with age, at least from middle-age onwards. This basic finding is confirmed in both multi-country cross-sectional studies (e.g., Hayward & Krause, Reference Hayward and Krause2015; Shulgin et al., Reference Shulgin, Zinkina and Korotayev2019), as well as in (mostly US-focused) longitudinal studies (e.g., Dillon & Wink, Reference Dillon and Wink2007; Bengtson et al., Reference Bengtson, Silverstein, Putney and Harris2015). However, in common with many hard-won scientific generalisations, it admits both significant variation and a number of exceptions. Indeed, although Silverstein and Bengtson’s (Reference Silverstein and Bengtson2018b) LSOG study did find evidence for a return to religiosity (i.e., a specific case of increased religiosity with age) among Baby Boomers in later life, this only applied to a ‘significant minority’ (p. 7). Hayward and Krause’s age-period-cohort analysis of data from eighty countries, using five waves of the World Values Survey, is worth quoting in this regard:

Such findings strongly suggest that different religions (or the same religion ‘inculturated’ differently in different places) are more or less susceptible to age effects. For example, Hu and Yang’s (Reference Hu and Yang2014) study of Chinese folk religion in Taiwan found a slight negative relationship between age and religiosity when controlling for cohort and period effects, which they plausibly ascribe to folk religion simply working differently from the types of religion dominant elsewhere (and not least the ‘elsewhere’ – i.e., the USA and western Europe – that predominate in the social-scientific literature). Relatedly, also worth highlighting is the fact that age effects do not apply uniformly across different dimensions of religiosity. For instance, Bengtson et al.’s reported ‘overall aging trend’ within their American LSOG data combines ‘an upward drift in subjective religiosity over the adult life course’ with ‘stability in religious service attendance’ (Reference Bengtson, Silverstein, Putney and Harris2015, p. 377). Variations in dimensionality would also help to explain cross-cultural differences in age effects, since different religious traditions tend to emphasise different styles of being religious.

This ‘widespread, but not universal’ positive association between age and religiosity is specifically a statement about adults – or at least, fully emerged adults. There exists a similar consensus of a negative association between age and religiosity from adolescence through emerging adulthood (Hardy & Nelson, Reference Hardy, Nelson, Crockett, Carlo and Schulenberg2022; see Section 3.4). However, it is worth noting that this is based on significantly fewer studies and relies much more heavily on longitudinal research in the USA; some studies hint that there may well be important cross-cultural differences here too (e.g., French et al., Reference French, Christ, Lu and Purwono2014, on Indonesian Muslim teenagers). While we focus on the adolescent-to-adulthood transition later, here let us note that these two big trends – that is, early-adult decrease and late-adult increase – combine to produce a shallow, roughly (and asymmetrically) U-shaped curve over the entire adult life course (Dillon & Wink, Reference Dillon and Wink2007).

3.2 Period Effects

In 430 BCE, an epidemic struck Athens, killing tens of thousands over a number of years. This devastation, coinciding with the early years of the Peloponnesian War, reportedly had a lasting impact on the Athenians’ religious belief and practice: ‘their supplications at sanctuaries and recourse to prophecies and the like were all of no avail. In the end they abandoned these, vanquished by the disaster… [P]eople, seeing nothing they could do as the disaster overwhelmed them, developed indifference toward sacred and profane alike’ (Thucydides, The Peloponnesian War, quoted in Ford, Reference Ford2023, p. 221).

This is a classic case of a period effect on religiosity, that is, a specific historical event or moment (in this case, lasting several years) significantly affecting one or more dimensions of religiosity. If age-period-cohort studies had existed in classical Greece, then there should have been a discernible ‘before’ and ‘after’ impact on various religious indicators, cutting across different age groups (though not necessarily impacting them in a uniform way, since it would presumably interact/interfere with different age- and cohort-related factors).

Some period effects are readily identifiable. Several major policy shifts in China over the past seventy-five years – not least Chairman Mao’s total ban on all religions during the Cultural Revolution and Deng Xiaoping’s significant ‘Reform and Opening Up’ from 1982 – have had clear impacts on religious life (Yang, Reference Yang2012). In Japan, the devastation of the Second World War and the end of State Shintoism brought with it a religious boom and proliferation of NRMs. Others are more difficult to discern. For example, in the USA and much of Western Europe, the 1960s to early 1970s appears to mark quite a significant religious watershed in Roman Catholic fortunes: church attendance and various other indicators of pastoral health start declining. But why? An awful lot happened both in the Church (Vatican II, significant liturgical change, Pope Paul VI’s encyclical Humanae Vitae) and the wider world (sexual revolution, counterculture, Vietnam, the Baby Boomers coming of age) at precisely that time. Which, exactly, is the cause of the observed period effect, is a hotly contested issue. Most likely, it is a combination of several things, all happening at once, and interacting with each other (Bullivant, Reference Bullivant2019).

In practice, a lot of period effects are like that. It is often easier to see that something happened at a particular time, or over a particular period, than it is to identify precisely what. Drawing clear lines between specific historical happenings and their effects on religious identities, beliefs, and behaviours often relies both on a good deal of implicit or explicit theorising. This is clearly true in historical work on religion. For example, attempts to explain major religious changes of the past must give cogent reasons why Big Events ‘X’ or ‘Y’ should have produced them. (Thucydides, for example, claims that it was the ineffectiveness of traditional devotional practices in protecting Athenians from the plague that undermined confidence in the gods themselves.) But this does not apply only to historians. For example, age-period-cohort studies of religiosity globally will henceforth need to account for the likely quite significant period effect of the COVID-19 pandemic. The short-term impact – clearly evident in churches’ attendance and other pastoral statistics from 2019 to 2022, for example – is relatively straightforward to understand (deaths, religious building closures, loss of habit). But longer term ones may well require much more by way of interpretation and speculation, especially as they are likely to vary very widely from place to place and religion to religion.

3.3 Cohort Effects

To understand what researchers mean by cohorts, it is easiest to start with big ‘generations’, familiar from such labels as ‘Baby Boomer’, ‘Millennial’, ‘Gen Z’, and so forth. The idea that such very large, generic groupings of people – essentially, everyone born within a specific fifteen- or twenty-year period – are helpful and meaningful categories is fairly well-established, and not only at the levels of pop culture caricature or journalistic shorthand. And with good reason. Obviously, there is a huge degree of internal variation – how could there not be? – but members of different generations do tend to differ in systematic ways from each other in terms of, say, moral or political outlook, or indeed religiosity. Nor is this simply a matter of relative age and ‘position’ within the population at a given time. Our Gen-Z students are different to how we were as Millennial students two decades ago. And we are both surely even more different from how our own Boomer and Gen-X professors were when they were undergraduates themselves (however hard that is to imagine).

Those brought up within a decade or so of each other spend their collective formative years with the same socio-cultural backdrop: big world events, economic conditions, cultural mores, what technologies are or are not available, popular music, movies, television, literature, and culinary trends. Importantly, they are also collectively raised by members of the previous generation (or two), using whatever parenting styles were prevalent among them. Having Boomer and/or Gen X parents (and indeed teachers, pastors/rabbis, etc.) is thus a large part of the Millennial experience. Needless to say, the upbringings of those born in a single generation are far from uniform. Differences of sex, class, race, immigration, local geography, and indeed religious background add up significantly. But it is still meaningful to think of all these sub-experiences overlapping, interacting, and adding up into a (very) broadly generalisable aggregated experience.Footnote ⁷

Although cohort-based analyses tend to deal with narrower birth-year bands, usually ten or five years, the essential idea here is the same: those born in 1975–1979 collectively have a somewhat different experience, from upbringing onwards, from those born in 1980–1984, who in turn have a different collective experience from those born in 1985–1989. Tracking different cohorts over time – for example, in multiple waves of cross-sectional surveys – reveals clear, systematic differences in religiosity. In many countries, especially in the West though by no means confined to it, one witnesses a clear stratification of religion over recent decades, with each cohort being progressively less religious than the previous one (e.g., Voas & Doebler, Reference Voas and Doebler2011; Shulgin et al., Reference Shulgin, Zinkina and Korotayev2019). That is to say, those born in 1985–1989 are less religious than the early '80s, who are less religious than the late '70s, and so on. This applies to different dimensions, such as religious affiliation, belief in God, and religious service attendance, albeit not uniformly (Stolz et al., Reference Stolz, de Graaf, Hackett and Antonietti2025). It is also independent of the age/life-course dynamics we noted earlier: each cohort may be more religious as adolescents and seniors than they are in middle-age, but they are a bit less religious than the preceding cohort’s adolescents and seniors were at the same ages and a bit more religious than the succeeding cohort’s are (or will be).

As ever, there are exceptions: not every cohort, in every country, on every dimension of religiosity, fits the basic generalisation. However, the overall trend is clear. Furthermore, the observed cohort effects are sufficiently strong to be the primary driver of significant religious decline in many countries. Large-scale secularisation is, then, predominantly driven by a gradual process of cohort replacement: ‘older, more religious people die and their less religious grandchildren come of age’ (Bruce & Voas, Reference Bruce and Voas2023). The intergenerational transmission of religion is, therefore, the key to understanding large-scale religious change over time. Better understanding how it works, and indeed why it seems not to work as effectively as it used to (since, obviously enough, the kind of cohort-by-cohort decline of recent history cannot always have been the case), is what the rest of this Element is about.

3.4 Religious Decline in Adolescence and Emerging Adulthood

If there is a point at which age and cohort effects meet, it is in adolescence and emerging adulthood (more on which in Section 4). Prior to adolescence, it is more difficult to say that children have their own religious identities, beliefs, and behaviours, even if some significant rites of passage occur in childhood (e.g., first confession and first communion for Roman Catholics, around age 7 years). Typically, as a matter of course, children’s religious identities are decided by their parents and caregivers, and their religious behaviours – both at home and in public – depend in large part on the adults who, for example, bring them to religious services. Children are also rarely asked, by social and psychological scientists, about their own religious views (though see Strhan & Shillitoe, Reference Strhan and Shillitoe2025; and Section 5.2). The large-scale surveys covered in this Element generally start asking children themselves questions in adolescence: NSYR begins at 13 years, ALSPAC and LSOG at 16 years, for example. For these reasons, questions about the intergenerational transmission of religion are really questions about whether or not children retain their parents’ religiosity as adults, that is, beyond childhood. And so, religious de-identification between childhood and adulthood – and thus, in adolescence – counts as intergenerational religious decline.

There is plenty of evidence that adolescence and emerging adulthood are significant periods of religious change in general, and religious decline in particular. Data from NSYR reveals that between Wave 1 in 2002/2003 (13–17 years) and Wave 4 (23–28 years) a decade later, the net change in religious identification was negative for almost all religious denominations and strongly positive (24%) for identification as ‘not religious’ (Denton & Flory, Reference Denton and Flory2020). In this study, losses were greatest among conservative Protestants (11.5%) and Roman Catholics (11.3%), and smallest for Jews (0.5%) and Mormons (0.9%). Both five-year periods between Waves 1 and 3, and 3 and 4 see similar levels of religious de-identification, but there is somewhat more decline in the earlier period than the older one, that is, prior to participants’ mid-twenties. Dyer et al. (Reference Dyer, Hardy and Goodman2022) provide a more fine-grained picture from ten years’ worth of the annual longitudinal Flourishing Families project: here, frequency of prayer and religious salience/importance decreased linearly from when participants were 12 years old; religious attendance declined more slowly at first, from around monthly, declining more rapidly in later adolescence to just a few or several times a year by age 19. The fourth generation of LSOG participants, recruited from age 16 in 1991, also showed decline from the start: here, various aspects of religiosity declined from age 16 through the twenties, before respondents began returning to religion in their thirties (Bengtson et al., Reference Bengtson, Silverstein, Putney and Harris2015). Similarly, Jones et al (Reference Jones, Cox, Cooper and Lienesch2016) report, based on Public Research on Religion Institute survey data, that 62% of religiously unaffiliated respondents who were raised in a religion claim to have left prior to turning 18 years old; 28% between 18 and 29 years; then very few after that. In their study of atheists – recruited via atheist, secular, and freethought organisations – Langston et al. (Reference Langston, Speed and Coleman III2020) found that the average age of onset of atheism was around 24 years, though there was a lot of variation (SD = 10.2).

All these findings are from the USA, but 2023/2024 Pew Research Center data from 36 countriesFootnote ⁸ paints a similar picture elsewhere too. Lesage et al. (Reference Lesage, Starr and Miner2025) did not sample under-18s, nor did they ask participants about the age at which they switched,Footnote ⁹ but they did compare religious switching rates of different age cohorts and found that the proportion of respondents in the youngest cohort (18–34 years) had already switched out of their childhood religion to the same or even greater extent than those from older generations. More people from this generation may continue to switch away from their childhood religion as they age; but we can see that a great deal of religious change has already taken place before their mid-thirties. At least some of this counts as intergenerational change; this is certainly so if people do not return to religion in later life to the same extent that they depart from it in youth, which they don’t.

4.1 Societal Factors in the Intergenerational Transmission of Religion

The intergenerational transmission of religion may be made easier or more difficult by various features of the societal (e.g., national) context. Indeed, it is plausible that secularisation itself makes it more difficult for parents to hand down their faith to their children. According to this hypothesis, the secular societal context undermines parents’ attempts to inculcate their children with their religious beliefs and practices. That is, secularisation is not only a product of the failure of intergenerational religious transmission, but also its cause. On the other hand, parents in secular contexts might be more motivated to make greater effort to counteract societal influences. Of course, this is not to say that their effort will be met with success.

Using 1991 ISSP data, Kelley and De Graaf (Reference 74Kelley and De Graaf1997) looked at the effects of parental religiosity in fifteen historically Christian territoriesFootnote ¹⁰ that varied on societal-level religiosity. Besides the obvious findings that (a) children of highly religious parents were themselves more religious than children of less religious parents, and (b) residents of highly religious countries were more religious than residents of less religious countries, Kelley and De Graaf (Reference 74Kelley and De Graaf1997) also observed an interaction. In highly religious countries, parental religiosity made relatively little difference; but in highly secular countries, parents play a strong role. For example, we can compare people whose parents attended religious services at least weekly and those whose parents attended religious services at most monthly. In very religious societies, both groups of people were relatively religious: having religious parents made some positive difference to people’s religiosity, but not a very big one. In contrast, in very secular societies, the gap widened, so that individuals from very religious families remained quite religious, whereas those from only moderately religious families had declined in religiosity. Further analyses of more recent ISSP data have replicated this finding (e.g., Müller et al., Reference Müller, De Graaf and Schmidt2014; Voas & Storm, Reference Voas and Storm2021; Ishii & Watanabe, Reference Ishii and Watanabe2024, all on data from up to 2008). However, the effect might not be robust against different measures and samples.

In their study of WVS data – covering over fifty territories – Gervais and Najle (Reference Gervais and Najle2015) found no evidence that the effect of religious upbringing (i.e., whether someone reported that they were raised to be religious) varied as a function of country religiosity. Furthermore, Voas and Storm (Reference Voas and Storm2021) have criticised Kelley and De Graaf’s (Reference 74Kelley and De Graaf1997) original study, which used parental religious behaviour (i.e., religious service attendance) as a predictor of their children’s religious belief. When they compared ‘like for like’ (p. 193), using religious behaviour for both measures, they found that parental and societal religiosity exerted independent effects, but did not interact in the way Kelley and De Graaf (Reference 74Kelley and De Graaf1997) and others observed. Furthermore, Voas and Storm’s (Reference Voas, Storm, Luijkx, Reeskens and Sieben2022) analysis of 2008 European Values Survey data – representing over forty territories – also found that parents in secular contexts were actually less likely to prioritise religious faith as a value for their children to learn at home, compared to those in more religious contexts: this was so regardless of how important religion was to the parents themselves.

Not everyone agrees with Voas and Storm’s (Reference Voas and Storm2021) insistence on comparing like with like. One reason to see if parental religious behaviours influence their children’s religious beliefs is to see whether children are more receptive to religious beliefs when they observe their parents behaving accordingly, that is, when they ‘walk the talk’. This point will be developed further in Section 5.2, when we discuss the role of ‘credibility-enhancing displays’ (CREDs; Henrich, Reference Henrich2009; Lanman & Buhrmester, Reference Lanman and Buhrmester2017). Similarly, Voas and Storm’s (Reference Voas, Storm, Luijkx, Reeskens and Sieben2022) results from their EVS study are at least in tension with research on religious minorities, which we shall discuss in Section 6.1. Overall, we can say that there is some evidence that parental influence on children’s subsequent religiosity depends on societal levels of religiosity, such that intergenerational transmission is easier in more religious contexts, but that the parental role is more important in more secular contexts. This latter claim can be understood in two complementary ways: first, that the onus rests especially heavily on parents in secular contexts; and second, that in such contexts parents can exert a greater effective role. However, we note that this effect of national religiosity/secularity is not very robust, as it varies depending on the dataset, and may be highly dependent on the measure of religiosity under examination.

There is quite a lot of research on how other societal-level variables affect religiosity in general, the most prominent of which is associated with Norris and Inglehart’s (Reference Norris and Inglehart2011; Inglehart, Reference Inglehart2021) existential security theory of secularisation. According to this theory, people seek out the comforts of religion when they experience existential threats (i.e., to their lives, health, livelihood) and, concomitantly, that they drift away from religion when such threats are absent. However, very little of this work examines intergenerational transmission specifically. Müller et al.’s (Reference Müller, De Graaf and Schmidt2014) research on ISSP data is among the rare exceptions., and found that parents have a weaker role in counties that suffer from high inequality; this may be because, to Norris and Inglehart’s (Reference Norris and Inglehart2011) point, individuals there might seek out religion regardless of their upbringing. Indeed, their analyses found income inequality to affect intergenerational transmission more strongly than even national religiosity. More research is needed that brings these two literatures – on existential security theories of religion and the intergenerational transmission of religion – together.

Departing from societal-level factors, we now turn to the role that the more immediate social context plays in the intergenerational transmission of religion. It is certainly plausible that a family’s – and, in particular, children’s and adolescents’ – immediate social context is a more powerful determinant in the intergenerational transmission of religion than their broader societal context.

4.2 Social Factors in the Intergenerational Transmission of Religion: Peers

The research literature of religious socialisation has typically focussed on four overlapping sources of socialisation, namely the home (i.e., parents and other family members), non-kin peer groups (e.g., friends), religious communities (e.g., churches), and educational settings (e.g., schools), with some recent research adding mass and social media to this list. Much of the existing research compares the relative importance of these different sources of religious socialisation, rather than examining their interactions. In this context, parents are often called ‘primary’ agents of socialisation, with other sources and settings being ‘secondary’: this is a reflection of the widespread finding that parental influence is typically stronger than that of others, including, for example, friends and teachers. Nevertheless, it is important to understand these other social sources of religiosity.

The research on peer influence on religiosity has mainly been on adolescents because peer influence is generally strong in this developmental period (Laursen & Veenstra, Reference Laursen and Veenstra2021). Some psychologists even argue that adolescence is a period of reorienting away from parents and towards peers (Blakemore & Mills, Reference Blakemore and Mills2014). This extends into emerging adulthood too: Montgomery-Goodnough and Gallagher’s (Reference Montgomery-Goodnough, Gallagher, Nielsen and Plakhotnik2007) review of research on the religiosity and spirituality of university-age adults found that nearly 80% of religious experiences and conversations were with friends (as distinct from family). Furthermore, as noted in Section 3.4, adolescence has also been found to be a period of religious decline. That said, it is important not to overly essentialise or reify adolescence and its relationship to religiosity. Adolescent religious trajectories are best characterised as diverse, varying as a function of various demographic factors, including gender, ethnicity, religious tradition, and national context (Hayward & Pearce, Reference Hayward and Pearce2021; Schnitker et al., Reference Schnitker, Medenwaldt and Williams2021). The rest of this section covers research on more individual- and family-level social factors.

Although parents are often found to be adolescents’ most important religious influences in general (Fisherman, Reference Fisherman2011; Hardy & Taylor, Reference Hardy and Taylor2024; Hardy et al., Reference 72Hardy, Hatch, Nelson and Schwadel2022; Regnerus et al., Reference Regnerus, Smith and Smith2004; Smith & Snell, Reference Smith and Snell2009), the influence of peers has also sometimes been found to be of very similar (or, sometimes, even greater) magnitude than that of parents, especially in the USA. For example, Regnerus and Uecker’s (Reference Regnerus and Uecker2006; see also Regnerus et al., Reference Regnerus, Smith and Smith2004) analyses of the 1996–1997 US National Longitudinal Data of Adolescent Health (Add Health) found that peers – close friends and schoolmatesFootnote ¹¹ – influenced adolescents’ (13–18 years) change in religiosity about as much as parents did, but in different ways. Peers played a slightly larger role in changes to religious behaviour, while parents played a slightly larger role in changes to religious attitudes.

Several cross-sectional studies also suggest that peers are important in adolescent religious development. Schwartz’s (Reference Schwartz2006) study on adolescents (14–18 years) found that perceived faith support – that is, the extent to which they felt that someone had supported their faith, whether via conversation or by example – from friends was a better predictor of religious commitment than from parents. Similarly Tratner et al.’s (Reference 83Tratner, Sela, Lopes, Shackelford, McDonald, Weekes-Shackelford and Abed2020) study, asking adults to rate childhood peers’ and parents’ religious involvement in their lives, found that shared religious involvement with peers during childhood was the better predictor of adult religious service attendance. On the other hand, when Tratner et al. (Reference Tratner, Sela, Lopes, Ehrke, Weekes-Shackelford and Shackelford2017b) asked adults to recall their adolescent (13–18 years) interactions with their best friend about religion, they found no difference between religiously affiliated and unaffiliated people.

This latter finding raises the question about whether adolescent peers influence religious deconversion specifically, as distinct from changes in degrees of religiosity. Tratner et al.’s (Reference Tratner, Sela, Lopes, Ehrke, Weekes-Shackelford and Shackelford2017b) study implies not. However, Langston et al. (Reference Langston, Speed and Coleman III2020) asked the same questions of atheists specifically, and found that having best friends who ‘showed their faith by how they talked and acted’ (p. 54, see also Section 5.2) delayed the onset of atheism. Broadening beyond adolescents’ best friends to their friendship group more generally, we see more of an effect. For example, Corcoran and Scheitle’s (Reference Corcoran and Scheitle2025) analysis of 2005 to 2013 National Study of Youth and Religion data (Waves 2 and 4) did find that the proportion of nonreligious individuals among later adolescents’ (16–21 years) five closest friends predicted whether or not they had shifted from a religious or uncertain worldview to an atheistic worldview eight years later.

These studies suggest that peers play some role in whether adolescents will remain religious, though longitudinal studies comparing peer and parental influences generally find that the latter is stronger (Hardy et al., Reference 72Hardy, Hatch, Nelson and Schwadel2022; see also Hardy and Taylor, Reference Hardy and Taylor2024). All the same, it would be worth looking more closely at the processes through which peers influence religiosity. This requires more research attention, as very little work has yet been done. Brechwald and Prinstein’s (Reference Brechwald and Prinstein2011) influential review of research on peer influence processes in adolescence failed to mention religion at all. Most of the research on peer influence assumes a conformity effect, such that adolescents are likely to be influenced in the direction of the majority/plurality of their peers. This is reflected in the common use of the average religiosity amongst peers as the predictor. Some studies have also considered whether religiosity’s association with popularity matters, finding that it does somewhat (Barrett et al., Reference Barrett, Pearson, Muller and Frank2007). But none of this provides very much detail on how adolescents influence one anothers’ religiosity.

Things have improved slightly in recent years, with more studies at least attempting to measure different ways in which peers interact around religion. Tratner et al.’s (Reference Tratner, Sela, Lopes, Ehrke, Weekes-Shackelford and Shackelford2017b, Reference 83Tratner, Sela, Lopes, Shackelford, McDonald, Weekes-Shackelford and Abed2020) work mentioned earlier uses their Childhood Religious Experiences with a Peer Inventory, which includes a variety of items about proselytisation, dialogue, and shared religious activities. Longitudinal studies like the NSYR and the Family Foundation of Youth Development project (Foundations) also include some relevant items. No clear conclusions can yet be drawn, however. For example, Tratner et al.’s (Reference 83Tratner, Sela, Lopes, Shackelford, McDonald, Weekes-Shackelford and Abed2020) research suggests that it is shared activities with peers (e.g., going to church together) that are important, rather than talking with peers about religion. And Hardy et al.’s (Reference 72Hardy, Hatch, Nelson and Schwadel2022) analysis of NSYR data seems to corroborate this, or at least suggests that religious conversations do not contribute anything over average peer religiosity more generally. On the other hand, Hardy and Taylor’s (Reference Hardy and Taylor2024) study using Foundations data does provide some evidence that peer encouragementFootnote ¹² is a better predictor of subsequent religiosity than peer religious involvement more generally.

Besides the specific processes involved in peer influence, more attention should also be paid to how different aspects of religiosity are influenced. For example, perhaps parental influence might be more important for determining whether a child grows up to be a religious adult, in terms of identity and belief; but peer influence might affect how actively and outwardly religious they are. Regnerus et al.’s (Reference Regnerus, Smith and Smith2004) work suggests something like this. Research on Muslim and Christian adolescents in Germany also bears this out: peer effects were found for religious service attendance, rather than for religious identification (Leszczensky & Pink, Reference Leszczensky and Pink2020).

The timescale of peer influence is also worth further investigation. Although adolescence is often described as a period during which peer influence eclipses parental influence, this is not to say that peers’ influences during adolescence will continue to make more of a difference than parents’ prior and abiding influences. Resistance against peer influence is itself subject to development, and Steinberg and Monahan (Reference Steinberg and Monahan2007) found that the ability to resist peer pressure grows linearly between ages 14 years and 18 years, whereas it was stagnant before and after that. It is therefore possible that peer influence from childhood, adolescence, and emerging adulthood will wane over time, while one’s religious upbringing – and parental influences more generally – may have a longer term effect. We will consider this in Section 5.

Clearly, there is more to discover about the role of peers during childhood, adolescence, and emerging adulthood in determining adult religiosity. Questions about the longevity of peer influences remain, as do questions about the mechanisms of influence and their effects on different aspects of religiosity.

4.3 Social Factors: Education

Having now considered the role of peers, including friends at school, it is worth also exploring the role of education and educational institutions more broadly, besides being a source of influential friendships.

According to one version of the secularisation thesis that modernization leads to the decline of religion, education – and especially scientific education, which is emphasised in modern syllabi – has a strong role to play. However, the empirical data show a more complex picture than this theory suggests. Schwadel’s (Reference Schwadel2015) correlational analysis of 2008 ISSP data from thirty-nine countries found that, averaging across countries, there was a negative relationship between university education and religiosity.Footnote ¹³ However, the pattern varied markedly between countries: eighteen of the thirty-nine (46%) countries showed the negative effect and twelve showed no significant effect, but nine actually showed a positive effect, such that more highly educated people were more religious. Furthermore, Schwadel (Reference Schwadel2015) found no relationship between country-level education (i.e., prevalence of university education in a country) and country-level religiosity (i.e., national average score on a composite religiosity measure and prevalence of ‘no religion’). Nor did the national average level of education affect the relationship between educational attainment and religiosity at the individual level; however, education’s negative effect on individual religiosity was stronger in highly religious countries than in more secular ones.

This is not to deny that there is an effect of education on religiosity and intergenerational religious transmission – after all, among Schwadel’s (Reference Schwadel2015) findings was an overall negative effect of educational attainment and religiosity at an individual level. This finding has also been replicated using 2008–2020 WVS and EVS data (Kavadias et al., Reference Kavadias, Spruyt and Kuppens2024). But the effects are subtler and more variable than is sometimes asserted or assumed. The variability is partly a matter of differences between countries, but also differences between religions and between aspects of religiosity, as we saw with peer influences in Section 4.2. Kavadias et al.’s (Reference Kavadias, Spruyt and Kuppens2024) WVS/EVS study – representing seventy-six countries – is bullish on the negative effect of educationFootnote ¹⁴ on religion, but also found exceptions. For example, at an individual level, educational attainment negatively impacted various aspects of religiosity,Footnote ¹⁵ but not religious service attendance. Furthermore, while they found national-level effects on various other aspects of religiosity, there was only weak evidence that people in poorly educated countries were also more likely to identify as religious (Appendix A8 of Kavadias et al., Reference Kavadias, Spruyt and Kuppens2024). They also detected some other cultural variation in their data: for example, in low-schooled Buddhist and Muslim contexts, individual-level education was positively associated with religiosity.

Neither of these studies is about the intergenerational transmission of religion per se, though their relevance is obvious. Studies on religious switching – that is, changing away from one’s childhood religion to something else (very often, but not invariably, to no religion) – get us closer. Reflecting the theme of the subtlety and variability of the effect of education on religiosity, Pew Research Center recently reported that education makes little difference to religious switching in most countries; they found that religious switching rates differed by educational level (having a post-secondary, e.g., university, qualification or not) in only twelve of the thirty-six countries sampled:Footnote ¹⁶ Brazil, Chile, Colombia, Greece, Italy, Japan, Netherlands, Peru, Poland, Singapore, Turkey, and USA (Lesage et al., Reference Lesage, Starr and Miner2025; see full list of countries sampled in Section 3.4, Footnote 8). Lesage et al. (Reference Lesage, Starr and Miner2025) did not speculate on why these countries differed from the others; nor do we discern an obvious pattern; these countries vary on religion as well as education. Focussing on the US context, Leon and Van Liew’s (Reference Leon and Van Liew2017) analysis of Baylor University’s 2005 Religion Survey might also permit us to make some inferences on intergenerational transmission, as it looks at the match between the respondent’s religious identity and those of their parents. They found no main effect of education: that is, highly educated people were no more likely to disaffiliate than anyone else. Nor did education predict changes in religious service attendance between age 12 years and adulthood (average age was ~54 years). Looking more closely at the data, they found that higher education did have a weak – and non-statistically significant – negative effect for Christians’ self-identification, but also a positive effect among those who grew up in non-Christian religions (i.e., in minority religions).

Overall, these studies do suggest a negative relationship between education and religiosity, but also that this relationship is highly variable, and depends on a variety of contextual factors. There has also been some suggestion – expressed by Leon and Van Liew (Reference Leon and Van Liew2017) and Schwadel (Reference Schwadel2015), among others – of the possibility that the effect of education is not on religion per se, as secularisation theorists suppose, but on conformity and innovation. According to this hypothesis, education enables and encourages people to hold views distinct from those normal and normative in the surrounding culture. The positive effect of education on minority religions is taken as evidence of this. If so, then we might expect that education may have positive effects on historically dominant religions too, as countries become more secular. Indeed, Schwadel (Reference Schwadel2015) mentions that among the nine countries where education has a positive relationship with religiosity are Sweden, New Zealand, and South Korea, all highly secularised countries.

Much of the research in this area focusses on university education as the difference-maker. This assumes that there is something important about this particular context – as distinct from secondary school education, for example – erhaps as one in which new ideas are explored and, concomitantly, old ones abandoned, including religious ideas imbibed at home during childhood. As we have already seen, the correlational research lends some credibility to this idea, but also complicates it. There has been some longitudinal research that tracks individual university students across their undergraduate careers, to see how their religiosity and spirituality changes. The research in the 1960s and 1970s, mainly on North American samples, suggested a secularising – or at least a liberalising – effect, but it did not take long for this picture to be problematised too (e.g., Hunsberger, Reference Hunsberger1978; Mayrl & Uecker, Reference Mayrl and Uecker2011). At the very least, the overall trend of religious decline tends to be qualified by inconsistencies across dimensions or aspects of religiosity, and contextual effects, including religious traditions and family-based characteristics.

Thus, Schwadel’s (Reference Schwadel2016) study of US individuals using 2002/2003–2027/2028 NSYR data found that, by the end of their undergraduate studies, participants reported weaker religious beliefs and less frequent private prayer, but neither less religious service attendance nor a weaker sense of personal religiosity (e.g., importance of religion, feelings of closeness to God).Footnote ¹⁷ In a follow-up study, Schwadel (Reference Schwadel2017) also found that this decline was strongest for mainline Protestants. Among the religiously unaffiliated, university education even had the effect of increasing religiosity, whose magnitude was at least that of the decline for mainline Protestants. According to this study at least – and consistent with the education and conformity/innovation view described earlier – university education seems to undermine the intergenerational transmission of both religiosity and irreligiosity. Bacon and Wang’s (Reference Bacon and Wang2024) analyses of National Longitudinal Study of Youth (NLSY97) data also found religious decline between young adulthood (18–22 years in 2002) and fifteen years later (33–37 years in 2017), attributable to university education; and that fundamentalistFootnote ¹⁸ Protestants, and to a lesser extent, fundamentalist Roman Catholics were most immune to the negative effects of tertiary education. As in Schwadel’s (Reference Schwadel2017) study, Bacon and Wang (Reference Bacon and Wang2024) also saw that university seemed to increase religiosity among previously religiously unaffiliated young adults.

Institutional characteristics might be expected to make a difference in whether students remain in the faith of their upbringing. Hill (Reference Hill2009) has compared the religious trajectories from before university to after graduation, of individuals enrolled at different types of US universities – public, private, and a variety of Christian universities – and found that besides the selection effects (i.e., highly religious individuals self-select into universities with strong religious identities, e.g., evangelical, Mormon), religious trajectories also differed across types of university. For example, there was a strong decline in religious service attendance among students at Roman Catholic and mainline Protestant universities, such that by the third year of university, average attendance was on par with public and private universities. In contrast, student religiosity was much more stable at conservative evangelical and Mormon universities. Hill (Reference Hill2009) also noted that compulsory chapel attendance predicted religious retention. Davignon and Thomson Jr (Reference Davignon and Thomson2015) then ran a follow-up study, focussed on universities associated with the Council of Christian Colleges and Universities (CCCU), which Hill (Reference Hill2009) categorised as ‘conservative evangelical’ and at which there was relatively little religious decline. They wanted to know what it was about these universities that helped to maintain student religiosity. Consistent with the research on peer influence, they found that the proportion of highly religious Christians (characterised here as ‘born again’ and ‘Bible-believing’) accounted for much of the effect. In contrast, denominational homogeneity made little difference. The perceived availability of ‘spiritual mentorship’ and ‘integration of faith and learning’ was also associated with various aspects of student religiosity, though it may also be that highly religious students are more likely to experience and perceive these differently than less religious students.

Some smaller studies, homing in on particular universities, also indicate another possible virtue of overtly religious educational institutions. Compare, for example, Brown et al.’s (Reference Brown, Blankholm, Hoesly, Hwang, Harris, Silverstein, Silverstein, Gärtner and Brown2024) study of an unnamed Protestant university in the American South against Stoppa and Lefkowitz’s (Reference Stoppa and Lefkowitz2010) study of a public university in the American Northeast. Both studies found significant religious decline – around 40% of the students at the Protestant university decreased in their religious service attendance and 46% of those at the public university did – but at Brown et al.’s (Reference Brown, Blankholm, Hoesly, Hwang, Harris, Silverstein, Silverstein, Gärtner and Brown2024) Protestant university, 24% of participants increased in their religious service attendance, compared to only 10% at Stoppa and Lefkowitz’s (Reference Stoppa and Lefkowitz2010) public university. In other words, even at religious educational institutions where significant declines in religiosity occur, there may also be more opportunities for increased religiosity here than at secular institutions.

Having considered institutional differences among universities, it is worth also briefly considering the difference that schools might make before students arrive at university, not least because adolescent declines in religiosity already predict young adult religious disaffiliation (Hardy et al., Reference Hardy, Hendricks, Nelson and Schwadel2023). McKendrick and Walker’s (Reference McKendrick and Walker2020) UK study is worth considering in some detail. Just under 19% of secondary schools in the UK are categorised as faith schools; most are Roman Catholic or Anglican, with other Christian, Jewish, Muslim, Sikh, and Hindu schools also represented. In contrast to the situation in the US, most UK faith schools are publicly funded, which partly accounts for their numerousness. UK faith schools are afforded a lot of flexibility on such matters as student entry and the appointment of teaching staff: in particular, they can impose religious requirements or exercise religious preferences. Of course, faith schools do vary in the extent to which they do so. McKendrick and Walker (Reference McKendrick and Walker2020) analysed longitudinal data from the Longitudinal Study of Young People in England (LSYPE; now known as the Next Steps study) that began in 2004, with students aged 13/14 years. McKendrick and Walker (Reference McKendrick and Walker2020) followed them until they were 25 years old and found that those who attended faith schools for their secondary education were much more likely to remain religious at age 25 years than those who did not. This effect was robust against other individual and contextual variables, including parental religiosity, and was independent of the effect of the child’s own religiosity during adolescence. In other words, this may be more than a selection effect, wherein highly religious parents send their children to faith schools, and the children may well have turned out to become highly religious adults anyway.

More research is needed on when and why faith schools have positive long-term effects on religiosity. As mentioned earlier, in the UK context, they have great flexibility in the selection of staff and the enrollment of students; they also have more flexibility than non-faith schools in their approach to religious education, and can elect to be more confessional and less impartial. All three of these factors may contribute to faith schools’ effects. Increasing the proportion of actively religious students would increase the peer effects discussed in Section 4.2. There is also some evidence that teachers can serve as religious role models and religious social influences more generally (e.g., Cohen-Malayev et al., Reference Cohen-Malayev, Schachter and Rich2014). Similarly, there is some evidence that confessional (as distinct from theologically ‘neutral’) religious education contributes to long-term religiosity (e.g., Arold et al., Reference Arold, Woessmann and Zierow2025). Of course, religious schools in different national, cultural, and denominational/religious contexts may differ from the UK situation in significant ways (see, for example, Uecker, Reference Uecker2009; Smith et al., Reference Smith, Longest, Hill and Christofferson2014; Wadsworth & Walker, Reference Wadsworth and Walker2017, in the American context). Furthermore, as with peers, the influences of teachers and religious education tend to be weaker than those of parents and the home (Bano & Ferra, Reference Bano and Ferra2018; Fisherman, Reference Fisherman2011; Francis et al., Reference Francis, Henchley and McKenna2025; Pusztai & Rosta, Reference Pusztai and Rosta2023). The effects of religious education can also at least partly be attributed to parents, insofar as they make decisions about their children’s religious education (see Section 5.1).

4.4 Social Factors: Religious Communities

Besides educational institutions, parents also influence children’s participation in other religious communities, especially those that congregate for worship and religious pedagogy, such as churches, synagogues, mosques, temples, gurdwaras, and so forth. The research on the impact of such communities on intergenerational religious transmission is surprisingly sparse, perhaps due to difficulties with operationalising and quantifying different congregational characteristics beyond the extent and frequency of childhood participation in congregational activities.

Certainly, childhood religious involvement predicts adult religiosity. American ten year olds who attend religious services at least once a week have been found to be more likely than others to maintain their frequent religious service attendance into their mid-twenties (Petts, Reference Petts2009; 1988–2004 data from NLSY79).Footnote ¹⁹ Even slightly less frequent (but still regular) attendees are more variable in their religious outcomes, with early decline likely to lead to significant decline by age 25 years. Unsurprisingly, children who rarely attend religious services as children very rarely do so as adults. Similarly, Hardy et al. (Reference Hardy, Hendricks, Nelson and Schwadel2023) NSYR study found that adolescents (13–17 years; 2002/2003) who attended religious services weekly were 32% less likely to have de-identified as religious a decade later (22–29 years; 2013) than those who attended just slightly less, two or three times a month. Taking a cross-sectional approach, Gemar’s (2023) analysis of 2018 GSS data found that adults who were more religious also recalled more frequent childhood religious attendance. Moon et al.’s (Reference 78Moon, Johnson, Case, Padgett, Johnson and Vander Weele2025) study was also cross-sectional but adds a cross-cultural dimension, using the first wave of the Global Flourishing Project: they likewise found that childhood religious participation was among the strongest and most cross-culturally robust predictors of religious belief, though this effect seems weaker in highly religious countries.

None of this is to say that childhood religious involvement guarantees the maintenance of religiosity as adults: some research finds that childhood and adolescent religious participation alone only weakly predicts adult religiosity, especially when other factors are considered (e.g., Krauss et al., Reference Krauss, Ismail, Suandi, Hamzah, Hamzah, Dahalan and Idris2013). This is partly because in highly religious countries, as the US was until quite recently, childhood religious service attendance was still quite common; the relative lack of variability makes it a poor predictor of future religious outcomes.Footnote ²⁰ Furthermore, childhood religious participation is a reflection, not only of the child’s own preferences, but also of the parents’, who have strong influences on children’s religious participation. On the other hand, encouraging their children to participate in communal religious life might be among parents’ most important strategies for maintaining children’s religiosity into adulthood.

Besides attending religious services, there are other ways in which children and adolescents engage with religious communities and institutions. We considered enrollment in religious schools and universities in Section 4.3. Religious communities also run forms of religious pedagogy aimed to supplement the mainstream education system; these Sunday schools, catechism classes, Hebrew schools, madrassas, and the like tend to occur at religious institutions (e.g., churches, synagogues, mosques) on weekends or during evenings. Youth groups also involve pedagogy, but tend to be less formal still, emphasising intrareligious socialisation and worship. Children and adolescents also encounter other adults in their religious communities, who might serve as their religious teachers, mentors, and role models. Similarly, religious communities are a source of peer influence (Section 4.2). As with religious service attendance, children’s participation in all these other activities is in part determined by parents; but it is nevertheless worth briefly considering their efficacy at maintaining religiosity into adulthood.

Research has consistently found that, of the ways in which adolescents can engage with religious communities, participation in youth groups is the best predictor of adult religiosity. NSYR studies have found that 13 to 17 year olds (Waves 1; 2002/2003) who attended a youth group had more positive attitudes towards religion three years later (Wave 2), and also greater intentions to continue attending religious services (Snell, Reference Snell2009); these effects tended to remain even when controlling for prior religiosity and other relevant variables. Youth group attendance also continued to predict religious service attendance in Wave 3 (2007/2008), at least among Protestants. This effect held even after controlling for various demographic and contextual variables, including parental religiosity, the child’s own prior religiosity, religious education, and Sunday school attendance (Vaidaynathan, Reference Vaidyanathan2011). While not always the strongest predictor, youth group attendance was among the most consistent predictors of adult religiosity.Footnote ²¹

These findings from NSYR data are echoed also in O’Connor et al.’s (Reference O’Connor, Hoge and Alexander2002) earlier and smaller study. The first wave was run in 1976, and mainly comprised 16 year olds. The investigators managed to track down and follow up with nearly half of the original participants when they were 38 years old in 1997/1998. They found that youth group participation was the most robust predictor of adult religiosity, even more so than parental religiosity, childhood church and Sunday school attendance, and other related variables.

It is difficult to say exactly why youth groups seem so effective a means for maintaining religiosity into adulthood, not least as youth group participation is rarely causally independent of other modes of engagement with a religious community, even if statistical techniques enable social scientists to estimate its unique contribution. It is possible, for example, that youth group participation is a good indicator of the confluence of the child’s and parent’s religiosity, and – in most cases – both have to consent to such participation, especially in the child’s later adolescence. It is also possible that youth group participation increases the number of religious peers in a child’s life; it may even contribute to increased religious homogamy, and having a religious spouse does increase one’s likelihood of remaining religious (McPhail, Reference McPhail2019; O’Connor et al., Reference O’Connor, Hoge and Alexander2002). Similarly, as Snell (Reference Snell2009) found, youth group attendance also increases the likelihood of having adult religious mentors and role models, who might contribute to long-term religiosity, perhaps especially in moments of doubt or other difficulty (though see Hardy et al. Reference Hardy, Amelang and Goodman2025; here, influence of other adults at church seems minimal). In other words, youth group participation might both reflect and contribute to a broader set of social resources available to children and adolescents that helps them maintain their religiosity.

4.5 Digital Media

We turn finally to the role that digital media – and the use of web-based technologies in particular – might play in the intergenerational transmission of religion. This is a relatively new field of research, and also an increasingly important one. The Internet provides broad access to religious (and anti-religious) ideas different from the ones encountered at home in childhood and adolescence. Social media platforms in particular can be sites of encounters with religious others that might be critical or even hostile to one’s religious faith. Furthermore, secular web-based activities might compete with private religious behaviours, as time is a limited resource. As such, Internet use might pose a potential threat to the intergenerational transmission of religion. On the other hand, the increased availability of religious content afforded by the Internet might serve to reinforce or enhance religiosity, as might increased social interactions with co-religionists. As with many technologies, real world effects depend on the circumstances of their use.

The data from the United States tends to suggest that Internet use has a deleterious effect on religiosity – with some caveats. Paul McClure’s work, using a variety of datasets, has been most influential on this topic; and the Baylor Religion Survey has been an important source of data here, which has also helpfully updated its social media questions over time. McClure’s (Reference McClure2017) correlational analysis of 2010 (Wave 3) Baylor Religion Survey data on US adults found that the amount of non-email Internet useFootnote ²² predicted declines in religious affiliation, though not religious service attendance and participation in other religious activities. McClure’s (Reference 77McClure2020) later analysis of 2017 (Wave 5) BRS data replicated the finding on disaffiliation, but this time also did find a negative relationship between Internet useFootnote ²³ and religious service attendance, as well as other aspects of religion, such as prayer, Scripture reading, and perceived importance of religion (though not spirituality).

It is possible that this increased negative effect of Internet use is driven by social media usage, which drastically increased between 2010 and 2017, itself in part aided by the rapidly increased adoption of smartphones. However, Laderi and Uecker’s (Reference Laderi and Uecker2025) recent analysis of 2021 (Wave 6) BRS data provides some reason for scepticism against blaming social media specifically:Footnote ²⁴ they found that, while social media use did not significantly predict any dimension of religiosity, playing online games was positively associated with agnosticism/atheism and negatively with a wide variety of attitudinal and behavioural religiosity indicators. Streaming video content (i.e., passively consuming content) was negatively associated with religious behaviours, but not religious attitudes. On the positive side, respondents who spent more time actively posting on social media websites also engaged in more religious behaviour (viz., religious service attendance and Scripture reading).

The BRS does not resample the same individuals with each wave, so it is not possible to run within-person longitudinal analyses on its data. It seems plausible that the positive relationship between social media posting and religiosity reflects an effect of religiosity on posting rather than the other way around: that is, religious people spend more time posting on social media. If so, then there is just no overall effect of social media use – whether active or passive – on religiosity. The BRS also samples US adults, which speaks less to the intergenerational transmission of religion than studies on children, adolescents, and young adults might. Uecker and McClure’s (Reference Uecker and McClure2023) study also relies on cross-sectional data, but the National Survey of Moral Formation (NSMF) from which it draws does focus on US adolescents (13–19 years old; 2017/2018). They found results consistent with Laderi and Uecker (Reference Laderi and Uecker2025). Adolescents exposed to more screen timeFootnote ²⁵ were more likely to say they were religiously unaffiliated; attended religious services less frequently; prayed less frequently; and perceived religious beliefs to be less important; this was so largely independent of parental religious affiliation, which means that screen time was negatively related to adolescents’ religion even when parents were highly religious. As with Laderi and McClure (Reference Laderi and Uecker2025), very few of these relationships seem to be driven by social media use per se. However, they did find some – albeit weak – evidence that adolescents in nonreligious families may use social media to explore and enhance their own religiosity. Both the BRS and NSMF are studies based in the US, and more work needs to be done to explore the relationship between Internet use and religiosity elsewhere. Gao et al.’s (Reference Gao, Xu, Tan and Yang2024) recent use of Chinese General Social Survey and WVS data is a good example of this: as in the US, social media use is also not much associated with religiosity in China.

Besides studies outside of the US, prospective longitudinal data on the effects of Internet use on religiosity is scarce. The NSYR did include questions about social media use, but the study was also run relatively early in social media’s explosive growth. In contrast to the previous studies, McClure’s (Reference McClure2016) analyses of NSYR data found that, compared to abstainers, young adults who used social media in Wave 3 (18–22 years; 2007/2008) then reported greater declines in religious service attendance five years later (22–29 years; 2013). Time spent on social media also had a liberalising effect, increasing endorsement of picking and choosing religious beliefs and even practising multiple religions. This latter finding may be interpreted as religious decline, but such religious reinterpretation may also allow individuals to remain within their religious faith. It is not clear why McClure’s (Reference McClure2016) results are at odds with the other studies described. It is possible that this is a function of its age, and much has changed about our Internet use in just a few years. On the other hand, its longitudinal design makes it worth taking seriously as a study in the causal relationship between Internet use and religiosity.

The positive potential of the Internet for intergenerational religious transmission – and religiosity more broadly – is still underexplored. It is well known that social media platforms increase the intensity of political opinion by making it easy for people to congregate with like-minded people and to view content consistent with their own attitudes (Van Bavel, Rathje, Harris, Robertson, & Sternisko, Reference Van Bavel, Rathje, Harris, Robertson and Sternisko2021); it might serve a similar function for religious beliefs, especially among those who deliberately seek out like-minded content and communities online. There is some empirical evidence that religious youths do turn to social media in this way. For example, Novak, Haselbacher, Mattes, and Limacher (Reference Novak, Haselbacher, Mattes and Limacher2022) interviewed religious adolescents and young adults (16–25 years) in Austria, who described discovering what Novak et al. (Reference Novak, Haselbacher, Mattes and Limacher2022) call ‘religious bubbles’, which the youths found to enhance their religious identities. Similarly, Haq and Kwok’s (Reference Haq and Kwok2024) interviews with adult Pakistani Muslims and Buddhists and Christians in Hong Kong found that respondents mainly entered ‘religious enclaves’ on social media, though they also described encountering stereotyping and other forms of hostility online. It is unclear how widespread this phenomenon is, and what its effects might be on the intergenerational transmission of religion. As noted earlier, it might help to reinforce inherited religious beliefs and practices; or it might lead to changes in religiosity, even when it does not lead to reductions in religiosity. As things stand, at least in the West, the use of the Internet is generally associated with religious decline, but there are also clearly opportunities for parents, religious leaders, and religious youths to use the Internet in ways that are more supportive of their religiosity.

5.1 Indirect Parental Influences on Children’s Religiosity

The ‘channeling hypothesis’ is the idea that parents’ main contributions to their children’s future religiosity are mainly indirect, via the decisions they make about their children’s social, educational, and specifically religious environments, which affect not only their exposure to religion, but also their choices of peers and other social influences. Himmelfarb’s (Reference Himmelfarb1979) seminal studyFootnote ²⁶ on the channeling hypothesis sampled over 1,000 Jewish adults in Chicago, and produced a causal model in which parental religiosity predicted children’s enrollment in religious schools; which then predicted their involvement in Jewish organisations during university; and all three initial factors then worked together to predict spousal religiosity, which had a powerful direct effect on determining participants’ current religiosity. Cornwall’s (Reference Cornwall and Thomas1988) study on nearly 2,000 Mormon adults was directly inspired by Himmelfarb’s (Reference Himmelfarb1979) study, and found very similar results: parents had relatively little direct impact on their children’s religiosity as adults, but did influence their adolescent religious observance and education, which predicted the religiosity of their peers both in adolescence and in young adulthood, which in turn predicted their religiosity as adults. Similarly, Willits and Crider’s (Reference Willits and Crider1989) study, based on longitudinal data collected in 1970 and 1981, found that the effect of parental religiosity during participants’ adolescence no longer predicted their religiosity as adults, once their own religious participation as adolescents was accounted for.

This early work on the channeling hypothesis should not be taken to mean that parental religiosity per se is insignificant in determining the adult religiosity of their children. Himmelfarb (Reference Himmelfarb1979) and Cornwall (Reference Cornwall and Thomas1988) themselves describe their findings as being about how parents’ decisions regarding their children’s social and educational contexts can support or, conversely, undermine intergenerational religious transmission. That is, the lesson is not that parental religiosity is unimportant, but that parents’ influence on their children’s involvement with religious environments, both formal and informal, is also very important. As we noted in greater detail in Section 4, among the decisions parents make on behalf of their children include the child’s participation in religious services (4.4), religious education (4.3), more informal settings like youth groups (4.4), which not only expose children to religious ideas and behaviours directly, but also increase the likelihood of their forming social relationships with peers and adults.

More recent research on the channeling hypothesis bolsters the general conclusion that indirect effects are important; but they also provide evidence for a more direct role for parents in passing religiosity to children. Martin et al.’s (Reference Martin, White and Perlman2003) cross-sectional study on US Protestant adolescents assessed parental, peer, and congregational influence simultaneously, and found that all three made independent contributions on adolescents’ religiosity (specifically, faith maturityFootnote ²⁷). They also found that parental influences partly went through peer influences but not congregational influences. There is some longitudinal data that is relevant too. Myers’s (Reference Myers1996) interviews with parents and their adult children already suggested that one’s parents’ religiosity during one’s childhood predicts one’s religiosity over a decade later, when the child – now an adult – has left home. Vaidyanathan’s (Reference Vaidyanathan2011) more recent analyses of three waves of NSYR (2002/2003–2007/2008) data on US adolescents and young adults also found that although some of the parental influence is explained by their role in determining their children’s social and educational contexts, parents’ own religiosity also exerts a significant independent role. Both their own participation in religious services and group activities as well as more private aspects of their personal religiosity predict their children’s religiosity five years later.

Vaidyanathan’s (Reference Vaidyanathan2011) results also included some differences across religious and cultural traditions, however. For example, although parental religious participation (e.g., church attendance) was more predictively powerful among White Protestants, Evangelicals, and Roman Catholics, it was more private or home-based aspects of parental religiosity that made a bigger difference among Black Protestants. Such differences should not be overstated. Certainly, the general finding that parental religiosity is among the best predictors of children’s religiosity seems generalisable across cultures. Similar findings have been reported in Malaysia, for example, where parental factors were found to be more important predictors of adolescent religiosity (age 16 years) than any others, including peer, school, and mosque involvement (Krauss et al., Reference Krauss, Hamzah, Ismail, Suandi, Hamzah, Dahalan and Idris2012, Reference Krauss, Ismail, Suandi, Hamzah, Hamzah, Dahalan and Idris2013). In sum, what parents do – at home and at church or mosque (and, by extension, synagogue, temple, etc.) – matters: it is not just about constructing the social and educational context.

5.2 Direct Effects of Parental Religiosity: Behaviours

In this subsection, we will consider the direct impact of parents’ own religious behaviour, especially as observable by their children, and also the extent and ways in which parents talk to their children about religion.

Among the things children learn from their parents are religious beliefs and behaviours. As we saw in Section 4.1, despite the decline in religiosity between generations, parental religiosity remains a significant predictor of children’s religiosity. Recall the analyses of ISSP data, covering up to forty-two countries, according to which parental religious service attendance when respondents were children (as they recalled) was a powerful predictor of those respondents’ own religious beliefs as adults, even in highly secular contexts, and perhaps, especially there (Ishii & Watanabe, Reference Ishii and Watanabe2024; Kelley & De Graaf, Reference 74Kelley and De Graaf1997; Voas & Storm, Reference Voas and Storm2021).

Cross-cultural longitudinal data is more scarce, and we rely again on NSYR data that has followed a panel of US adolescents (13–17 years) since 2002/2003. Analyses of these data consistently find that parents’ own religious involvement (e.g., their own religious service attendance) and the religious environment at home, which is usually instigated by parents (e.g., prayers during meals, family conversations about religion), are both among the strongest and most robust predictors of a child’s religiosity even after they leave home as emerging adults. It is a better predictor of future religiosity than, for example, the child’s own religious service attendance, Sunday school attendance, and youth group attendance (Vaidyanathan, Reference Vaidyanathan2011; pace Willits & Crider, Reference Willits and Crider1989). According to NSYR data, in the long term, parental influence is stronger and more robust than peer influence and church influence (Hardy et al., Reference 72Hardy, Hatch, Nelson and Schwadel2022; Stanford et al., Reference 82Stanford, Futris, Brown and Lemons2024; see also Hardy et al., 2024, using a different US dataset). Regnerus and Uecker’s (Reference Regnerus and Uecker2006) study of Add Health data is also worth revisiting (see Section 4.2). Although both peer and parental religiosity similarly predicted religious change/stability in adolescents, the parental influence was more robust and comprehensive: it affected both private and public aspects of religiosity and predicted both increases and (lack of) decreases in religiosity.Footnote ²⁸

As noted earlier in Section 4.2, there is some evidence that parental influence gives way to peer influence during adolescence, but the cross-sectional data on adults from ISSP, WVS, and other studies also provide evidence that parental religiosity exerts a long-term effect. Such studies rely on adults’ recollection of their parents’ prior religious behaviours, and it is possible that our present levels of religiosity affect the way we recall the past. Nevertheless, the consistency between longitudinal and cross-sectional analyses is reassuring. There is also both longitudinal and cross-sectional evidence that provides more detailed texture about the means through which parents might influence the religiosity of their children.

Recall Hardy et al.’s (Reference 72Hardy, Hatch, Nelson and Schwadel2022) findings from NSYR and Family Foundations of Youth Development longitudinal data that parental factors were stronger and more robust than peer factors (Section 4.2). Besides asking parents about their religious service attendance and how important their faith was to their daily lives, the NSYR also asked about how often the family prayed together before meals and how often they discussed religious matters. Among these, parental religious service attendance and family religious conversations (see Section 5.3) were most protective against children’s religious disaffiliation within the next five years. The Family Foundations study included other measures of family religiosity too, besides parental religious involvement. There was a measure of family religious activities (e.g., reading Scripture together, praying together outside of meals, singing religious songs together); a measure of parents’ attitudes towards their children’s religiosity (e.g., how important it is that they prayed); and a measure of parents’ efforts to pass on their faith (e.g., whether they teach their children about their faith). This time, parental religious attendance itself was relatively unimportant, relative to their attitudes towards their children’s religiosity and the frequency of family religious activities.

The literature on the importance of ‘credibility-enhancing displays’ (CREDs) also tends to use more specific items than just parental religious attendance. Put simply, a CRED is a ‘costly’ (e.g., in terms of time, effort, financial outlay) action accompanying a professed belief, indicating that the person really believes what they claim to believe (Henrich, Reference Henrich2009). Martyrdom is thus an extreme example of a religious CRED. But more prosaically, anything that demonstrates a willingness to ‘practise what you preach’ might count. Devoting hours of one’s life to praying or attending (not necessarily very exciting) services, foregoing the pleasures of alcohol or bacon, tithing a significant chunk of one’s income – these are all prima facie evidence that you really do believe in the truth of a religious system in which such sacrifices are rendered ‘worth it’. According to CREDs theory, such evidence makes the religious system itself more compelling too: if other people – including our parents – take a belief system seriously enough to engage in such costly behaviour, then maybe we should too.

Langston et al.’s (Reference Langston, Speed and Coleman III2020) study on atheists contained four measures of parental or family religiosity. They used Lanman and Buhrmester’s (Reference Lanman and Buhrmester2017) CREDs scale, which asked about parents’ religious service attendance, but also their religious volunteerism, the extent to which they were good role models, and other similar indicators that they took their faith seriously. Besides this, Langston et al. (Reference Langston, Speed and Coleman III2020) also included measures of how important religion was in the family during their childhood, whether parents allowed them to make their own religious choices, and whether there was conflict over religious differences between parents and children. Although they found that CREDs quite consistently delayed the onset of atheism, some of the other variables were actually more powerful predictors, especially participants’ sense of the overall importance of religion in the home. Conversely, those who strongly recalled their parents allowing them to make their own religious choices were earlier to adopt atheism. Religious conflict in the home also predicted early atheism; but conflict may well be a symptom of religious decline rather than a cause.

Besides predicting the age of onset of atheism, CREDs has also been found to predict adult religiosity more generally, such that people who witnessed more behaviours consistent with professed faith, especially from parents, were more likely to remain religious believers (Gervais et al., Reference 71Gervais, Najle and Caluori2021). Predating the theoretical work on CREDs, Bader and Desmond’s (Reference Bader and Desmond2006) study using Add Health data already found that adolescents whose parents’ religious attitudes and behaviours were aligned (e.g., they said that religion was of great personal importance and also attended religious services weekly or more) were themselves more religious: they attended religious services and prayed more frequently, said that religion was of greater importance to them, and agreed more strongly that their Scriptures were divinely inspired and inerrant. Lanman and Buhmester’s (Reference Lanman and Buhrmester2017, Study 1) original study on religious CREDs also included a religious emphasis scale (Hunsberger & Altemeyer, 2006), which asked about the extent to which religiosity was encouraged and emphasised (typically by parents) during childhood: both were correlated with adult religiosity, but CREDs had the stronger and more robust independent effect. In their second study, they also found that CREDs were a better predictor of current religiosity than participants’ religious upbringing per se. Besides evidence from the US, similar results on the importance of CREDs have been found in the Czech Republic and Slovakia (Willard & Cingl, Reference Willard and Cingl2017), Iran (Bagherian et al., Reference Bagherian, Sayyad, Sheibani and Rios2025), Ireland (Turpin & Willard, Reference Turpin and Willard2022), and Poland (Łowicki & Zajenkowski, Reference 76Łowicki and Zajenkowski2020). Maij et al. (Reference Maij, van Harreveld, Gervais, Schrag, Mohr and van Elk2017) described their study in the Netherlands, Switzerland, and the US in terms of CREDs, although their measure mainly comprised items about parental and family religiosity: still, it provides evidence that having observed religion being taken seriously at home, particularly by parents, is important for maintaining religiosity into adulthood.

All this can be taken as evidence for the importance of religious behaviours, typically instigated by or otherwise involving parents, which demonstrate to children that their parents take their religious commitments seriously and that those religious commitments are an important aspect of the family’s life. Thus, parents’ own religious service attendance, the active presence of religious practices at home (e.g., praying together, singing hymns), and to some extent, the discouragement of religious choice (e.g., Langston et al., Reference Langston, Speed and Coleman III2020; Tratner et al., Reference Tratner, Sela, Lopes, Ehrke, Weekes-Shackelford and Shackelford2017a) all contribute to children’s maintenance of their religiosity. Having looked at what parents do, we now turn to what they say about religion and how they say it.

5.3 Direct Effects of Parental Religiosity: Conversation

We have just seen that to transmit religion across generations, it is important for children to witness their parents ‘walking the talk’; but it is also important for talking to take place. Multiple studies have found that the frequency of family discussions about religion during childhood predicts adult religiosity. As noted above, analyses of NSYR data have found that teenagers from families that discussed religion more frequently were much more likely to remain religious (e.g., Hardy et al., Reference 72Hardy, Hatch, Nelson and Schwadel2022; Smith & Adamczyk, Reference Smith and Adamczyk2020; though also Hardy et al., Reference Hardy, Amelang and Goodman2025Footnote ²⁹). Smith’s (Reference Smith2021) analysis of NSYR data even found that differences in the frequency of family religious conversations and family prayer together could explain the religious attendance gaps between children of religious conservatives, moderates, and liberals. This finding about the importance of religious conversations has also been corroborated by Legault-Leclair’s (Reference Legault-Leclair2025) study on Canadians.

Regarding the content of these parent–child interactions, there is still relatively little research. Tratner et al.’s (Reference Tratner, Sela, Lopes, Ehrke, Weekes-Shackelford and Shackelford2017a, Reference Tratner, Sela, Lopes, Ehrke, Weekes-Shackelford and Shackelford2017b, Reference 83Tratner, Sela, Lopes, Shackelford, McDonald, Weekes-Shackelford and Abed2020) studies with their Childhood Religious Experiences Inventories are worth revisiting here (see Section 4.2). Recall that these studies with US participants, asked adults to retrospectively recall childhood experiences with religion, with their peers on one hand and with their caregivers (usually parents) on the other. Although childhood interactions with peers were not very predictive of whether adults were religious affiliated or unaffiliated (Tratner et al., Reference Tratner, Sela, Lopes, Ehrke, Weekes-Shackelford and Shackelford2017b), interactions with caregivers were more predictive. Consistent with Langston et al. (Reference Langston, Speed and Coleman III2020), those who were religiously unaffiliated were more likely to recall their parents permitting them to choose their religious beliefs and practice, and to explore other religious options. In contrast, adults who remained religiously affiliated were more likely to recall religious assurances and affirmations from caregivers (e.g., being told that God loves them, cares for their family). More negative parental interactions, like expressions of disapproval over their religious choices or comments about God’s disapproval of behaviour, seemed not to affect adult religiosity either way. This latter finding might suggest that although Langston et al. (Reference Langston, Speed and Coleman III2020) found that religious conflict between parents and children was associated with earlier onset of atheism, perhaps the relationship isn’t directly causal.

Besides content and frequency, the manner in which parents speak to their children about religion may also play a role here. Dollahite and Thatcher’s (Reference Dollahite and Thatcher2008) interviews with highly religious American families (including parents and children) across different religious traditions led them to conclude that youth-centred rather than parent-centred conversations were more likely to be effective, as they were more emotionally positive for both children and parents. Parent-centred conservations are typified by a unilateral approach, in which the parent does most of the talking, with little attention paid to the child’s needs, interests, and opinions. Youth-centred conversations not only make space for the children to speak, but are also to some extent instigated and led by the children; furthermore, there is concern to nurture the relationship between the parent and the child, not just to transmit information from the parent to the child. In other words, children – and, indeed, parents – enjoy conversations that are more dialogical, and less narrowly didactic. More longitudinal research is needed on whether these conversational styles actually predict different religious outcomes down the line.

At a finer grain of analysis, there is also increasing evidence for the importance of subtle verbal cues in how parents speak to their children about religious matters. This work has emerged from Paul Harris’s research on the role of testimony – including parental testimony – in children’s formation of beliefs about invisible entities, including both scientific as well as religious entities (e.g., Harris & Koenig, Reference Harris and Koenig2006). Over the past twenty years, Harris and his collaborators and others have established that children as well as adults across different cultures – including American, Mayan, Chinese, Iranian, and Turkish – often express more confidence in the existence of invisible entities posited by science (e.g., germs, oxygen) than in the existence of invisible entities posited by religion (e.g., God, souls).Footnote ³⁰ Only young preadolescent children sometimes expressed equal confidence in the existence of both kinds of beings; average confidence in religious entities never surpassed average confidence in scientific entities (see Ma et al., Reference Ma, Payir, McLoughlin and Harris2024 for details). They have also investigated the causes of this discrepancy, among which are subtle verbal cues that parents and other adults use in talking about scientific and religious phenomena, respectively.

The research on how parents talk to children about religion has found three kinds of subtle speech cues that might contribute to lowering children’s confidence in the existence of religious entities like God. First, parents use language that expresses uncertainty, using words like ‘maybe’, ‘kind of’, and ‘I guess’. Second, parents mention that there is variation in what people believe. For example, they might point out that not everybody believes in God, or that not all religions have a concept of heaven. Third, parents make what McLoughlin et al. (Reference McLoughlin, Jacob, Samrow and Corriveau2021) call ‘explicit claims about the ontological status’: that is, they explicitly mention that God ‘exists’ or that heaven is ‘real’. These speech cues were much more common when parents talked about religious phenomena than when they talked about scientific phenomena. Highly religious parents were less likely to use uncertainty laden language, compared to less religious parents; but even in their case, there was a slight difference between how they talk about religious versus scientific entities. Furthermore – and crucially – these verbal cues were correlated with children’s (~5–7 years old) confidence in the existence of these religious entities. These findings have also been partially replicated in Iran and China (McLoughlin et al., Reference McLoughlin, Cui, Davoodi, Payir, Clegg, Harris and Corriveau2023), with the presence of uncertainty cues when talking about religion being the most robust finding.

5.4 Effects of Parenting Approaches on Children’s Religiosity

We have already encountered a few studies that associate religious disaffiliation with parents giving their children more religious choices (Tratner et al., Reference 83Tratner, Sela, Lopes, Shackelford, McDonald, Weekes-Shackelford and Abed2020; Langston et al., Reference Langston, Speed and Coleman III2020; see also Thiessen & Wilkins-Laflamme, Reference Thiessen and Wilkins‐Laflamme2017). However, this is not to say that the most effective way for parents to transmit their religiosity to their children is to rigidly deny any agency to their children. Recent research makes clear that children themselves desire some autonomy in religious matters. We see this in their preferences for youth-centred religious conversations earlier, for example. Dollahite et al.’s (Reference Dollahite, Marks, Babcock, Barrow and Rose2019) study on highly religious families also found that a common approach is to balance firmness with flexibility, avoiding both rigidity and laxness (see also Barrow et al., Reference Barrow, Dollahite and Marks2021). Their conclusion is consistent with the work on CREDs described earlier: a degree of religious firmness demonstrates to children that religion is valuable, but this should be tempered with some flexibility, acknowledging children’s own preferences as well as life circumstances. After all, there is some evidence – for example from interviews in the LSOG – that although strong religiosity is much more likely to be passed onto the next generation than moderate and weak religiosity (Hwang et al., Reference Hwang, Yoon, Brown and Silverstein2024), strictness about religion can sometimes trigger a backlash from children and adolescents (Bengtson et al., Reference Bengtson, Putney and Harris2013). Here too more research is required on what sort of balance predicts what sort of religious trajectory. And of course, what that optimum balance might look like in any given context, or indeed with any given child, could vary rather significantly.

Research on parenting styles and parent–child relationships more broadly can also speak to this question. For example, in their study of Dutch adults who were asked to recall their parents’ parenting styles, Vermeer et al. (Reference Vermeer, Janssen and Scheepers2012) found that strict parenting style negatively predicted religious service attendance as adults. This is consistent with the LSOG finding just mentioned. Perhaps surprisingly, however, authoritative parenting style – usually construed as a balance between authoritarian and permissive styles (Larzelere et al., 2013) – was not predictive either way (see also Pinquart & Silbereisen, Reference Pinquart and Silbereisen2004). On the other hand, Hardy et al.’s (Reference Hardy, White, Zhang and Ruchty2011) US study did find that parents’ and children’s religiosity and spirituality were more closely matched among the parents who used an authoritative parenting style. This was regardless of parental religiosity: in other words, for both religious and nonreligious parents, authoritative parenting is helpful for intergenerational transmission.

The evidence for the role of emotionally positive relationships is less equivocal. This can be seen in LSOG data, according to which feelings of closeness with parents (as rated by the children) strongly predicts how similar their religious views are to their parents’; and this applies across religious traditions. A variety of other studies have also found that adolescents who enjoy emotionally positive relationships with their parents – measured in various ways, including warmth, affection, communicativeness, and trust – are more likely to remain religious (Desrosiers et al., Reference Desrosiers, Kelley and Miller2011; French et al., Reference French, Eisenberg, Sallquist, Purwono, Lu and Christ2013; Longo & Kim-Spoon, Reference Longo and Kim-Spoon2014; Łysiak et al., Reference Łysiak, Zarzycka and Puchalska-Wasyl2020; Power & McKinney, Reference Power and McKinney2013; Regnerus et al., Reference Regnerus, Smith and Smith2004). Bengtson et al.’s (Reference Bengtson, Putney and Harris2013) LSOG interviews with ‘prodigals’ – those who rebelled but then returned to the faith later on – also suggest that parental warmth is an important factor here. Taken together, it does seem that parenting style, and the parent–child relationships that ensue are clearly important for the intergenerational transmission of religion.

5.5 Family Structure and Children’s Religiosity

Besides parenting style, there are other features of the family that are predictive of children’s religiosity. The two that have received the most attention are the relationship status of the parents, and the extent to which they are religiously similar.

With regard to relationship status, studies generally find that children and adolescents who are brought up in families that include both of their biological parents are more likely to be religious than those who are brought up in other kinds of household, and perhaps especially single-parent households (Day et al., Reference 69Day, Jones-Sanpei, Smith Price, Orthner, Hair, Moore and Kaye2009; Uecker & Ellison, Reference Uecker and Ellison2012). Furthermore, it does not seem to be family structure per se that shapes children’s religious trajectory, but the experience of their parents’ divorce. For example, Denton’s (Reference Denton2012) NSYR study found that adolescents (13–17 years) whose parents were already divorced during Wave 1 were no less religious five years later than those whose parents were married, but those who had subsequently experienced their parents’ breakup ended up less religious later on (see also Lau & Wolfinger, Reference Lau and Wolfinger2011; Lawton & Bures, Reference Lawton and Bures2001; Zhai et al., Reference Zhai, Ellison, Glenn and Marquardt2007). There are various possible reasons for why divorce might affect religiosity, one of which is that it affects the religiosity of parents themselves and also the children’s exposure to and experience of their parents’ religiosity.

Parental religiosity and divorce are related in two ways. First, highly religious parents are less likely to get divorced (Mahoney, Reference Mahoney2021), and so children of divorced parents also happen to be, on average, children of less religious parents. In this case, it is not the effect of family structure per se that predicts decreased religiosity among the children; this is just a function of the weaker religiosity of the parents. Indeed, there is some evidence that the association between family structure and religiosity largely disappears once parental religiosity is accounted for (Uecker & Ellison, Reference Uecker and Ellison2012). On the other hand, as noted above, Denton’s (Reference Denton2012) results do suggest an effect of divorce per se on children. The other way in which parental religiosity and divorce are related is that divorce has a negative impact on religiosity for the divorcees themselves (Tanaka, Reference Tanaka2010). In this case, parents’ decreased religiosity might then be passed down to children. This view is consistent with Denton’s (Reference Denton2012) finding that adolescents’ and emerging adults’ decline in religiosity post-dates their parents’ divorce. Of course, divorce may also have other social psychological effects on the children relevant to religiosity, besides reducing parental religiosity.

Divorce reduces children’s exposure to religious role models in a straightforward way: depending on the custody arrangements, at least one parent typically becomes less involved in the children’s lives (e.g., instead of seeing both parents every day, they might only spend time with each parent for half the week). Given that parents play a strong role in their children’s religious socialisation, decreased exposure to parents would often lead to a reduction in this source of religious influence. Zhai et al. (Reference Zhai, Ellison, Glenn and Marquardt2007) has reported evidence of this, and specifically that the deficit post-divorce in fathers’ involvement in religious socialisation accounts for much of the decline in children’s religiosity. Similarly, Petts’s (Reference Petts2015) NSYR study found that although family structure was not directly predictive of religiosity, it did weaken the relationship between parents’ and children’s religiosity. This account of the effect of divorce on children’s religiosity is consistent also with findings about the religious composition of households and their effect on children’s religiosity.

Religious homogamy is the marriage between two people of the same or substantially similar religious backgrounds and/or dispositions. Religious heterogamy is the marriage between two people of different religious backgrounds and/or dispositions. Many studies have found that religiously heterogenous couples are less effective at passing down their religiosity. In a way, this seems almost inevitable. For example, if two parents belong to different religions, a child would usually have to choose one religion to adopt as their own (more often the mother’s; see Section 5.6), or they would have to attempt to blend them, which usually entails only partially observing either. Indeed, there is some evidence that couples in interfaith marriages are less religious (Pew Research Center, 2016, October 26), and that this may be more than a function of self-selection but also of social influence (Voas, Reference Voas2003). If so, then this decreased religiosity among parents would affect how much religious input their children receive at home. Indeed, Pew Research Center (2016, October 26) reported – based on their 2014/2015 US Religious Landscape Study – that those raised in religiously mixed (viz., Protestant–Catholic) households were less likely to be exposed to religious ideas and activities, whether at home or in religious communities.

Religiously heterogenous couples do seem to bring up less religious children; children of religiously heterogenous parents are also more likely to be disaffiliated as adults than those raised in religiously homogamous families (though see Bengtson & Silverstein, Reference Bengtson, Silverstein and Timonen2018, based on 2005 LSOG data). 14% of children with two Protestant parents and 19% of children with two Catholic parents ended up unaffiliated as adults; in contrast, 26% of children with one Protestant and one Catholic parent ended up unaffiliated (Pew Research Center, 2016, October 26. In their more recent 2019 Survey of Religion Among U.S. Teens and Their Parents, Pew Research Center (2020, September 10) again found similar, if even more drastic results: 44% of teens (ages 13–17 years) in religiously mixed homes were unaffiliated, compared to 29% whose parents share the same religious affiliation. Broadening the cultural scope, McPhail’s (Reference McPhail2019) study on 2008 ISSP data, covering thirty-seven countries, also found that respondents whose parents had different religious affiliations were themselves significantly less religious on various dimensions than those whose parents had the same religious affiliation. Based on his statistical mediation analysis, he also concludes that this is because parents in mixed marriages are themselves less religious than those in homogamous relationships. However, these kinds of statistical analyses should be interpreted with caution.

As we might expect, mixed marriages between religious and nonreligious individuals pose a particular problem for the intergenerational transmission of religion. The 2016 Pew Research Center study cited earlier reported that 34% of those raised in Protestant–Unaffiliated homes, and 42% raised in Catholic–Unaffiliated homes, identified as religiously unaffiliated, compared to 13/19% raised in fully Protestant/Catholic and 26% raised in mixed Protestant–Catholic homes.Footnote ³¹ McPhail’s (Reference McPhail2019) ISSP-based study certainly found this too: having one unaffiliated parent had a stronger effect than having parents from two different religious traditions. David Voas and colleagues have found that children who only have one religious parent are half as likely to be religious in adulthood than those who have two religious parents (Voas & Crockett, Reference Voas and Crockett2005 for the UK; Voas & Storm, Reference Voas and Storm2012 for England and Australia). It is worth mentioning here that the effect of religious heterogamy can cut both ways, and work against nonreligion: analysing Chinese data, McPhail and Yang (Reference McPhail and Yang2020) found that having one religious parent increased children’s religiosity in that predominantly secular context.

Besides differences in religious identity, gaps between parents in religiosity also weaken intergenerational transmission. In McPhail’s (Reference McPhail2019) study, children from households in which only one parent frequently attended religious services were less religious as adults than those from households in which both parents were frequent attendees. Similarly, Pew Research Center (2020, September 10) reported that teenagers who attended worship services with both their parents – implying, of course, that both parents attend – were substantially more likely to say that they shared ‘all the same religious beliefs’ as their parents than those who only attended with one parent, 61% versus 45% respectively.

5.6 Coda on Fathers versus Mothers, and Grandparents

There is a long-standing debate in the literature about the relative influence and importance of maternal versus paternal influence in the transmission of religiosity to children. The most common finding, certainly in the US, has been that mothers play a stronger role in the intergenerational transmission of religion. This was found in the Pew Research Center (2016, October 26) study cited earlier, for example: in mixed Protestant–Catholic families, people tended to adopt the religion of their mothers. Similar results have been reported in previous studies in the US, including among African American (Gutierrez et al., Reference Gutierrez, Goodwin, Kirkinis and Mattis2014) and Asian American (Park & Ecklund, Reference Park and Ecklund2007) families; and also in the UK (Francis & Brown, Reference Francis and Brown1991). Stearns and McKinney’s meta-analytic study of thirty studies published between 1990 and 2017 also found that mothers’ religiosity was a better predictor of intergenerational transmission than fathers’ religiosity. Furthermore, mothers’ influence was stronger than fathers’ for both sons and daughters, though the mother–daughter and father–son influences were stronger than the mother–son and father–daughter influences, respectively. The standard explanation for the strong role of mothers is that – at least in the Western cultural contexts under consideration – women tend to be more religious than men, and that mothers are more often responsible for religious socialisation than fathers, as Pew Research Center (2016, October 26) also reported.

It is important not to treat this finding as a universal and indelible fact. Research on the intergenerational transmission of other kinds of learnt traits may be instructive here. For example, in her study on political socialisation, Weiss (Reference Weiss2023) noted that although previous research found that fathers played a stronger role in the 1960s, more recent evidence suggests that maternal influence is now as strong as that of paternal influence (see also Boonen, Reference Boonen2017; Rico & Jennings, Reference Rico and Jennings2016). Indeed, there is some recent indication that an analogous shift – albeit in the opposite direction – may be taking place when it comes to the intergenerational transmission of religiosity: fathers may now be playing a stronger role than they once did.

Ishii and Watanabe’s (Reference Ishii and Watanabe2024) study of 2008 ISSP data from forty-three countries found no difference between maternal and paternal religious behaviour on respondents’ belief in God. There is even some evidence of a fathers’ roles becoming more important than mothers: in Conway et al.’s (Reference Conway, Spruyt and Coutinho2023) study on 2018 ISSP data from twelve European Catholic countries, fathers’ religious participation was a better predictor than mothers’ on several dimensions of religiosity: religious self-perception, prayer, church attendance, and belief in God. This chimes with a similar finding in Baker-Sperry’s (Reference Baker-Sperry2001) research on American Catholics. It is not just Catholics who should attend to fathers’ religiosity. Luria and Katz’s (Reference 76Łowicki and Zajenkowski2020) study on the transmission of religious and secular values in Israel finds comparable results among Israeli Orthodox Jews. For example, comparing religiously mixed families in which only the father or mother was Jewish, they found that those with Jewish fathers shared more religious values with them than those with Jewish mothers – this, despite the fact that Judaism is matrilineal. Luria and Katz (Reference Luria and Katz2020) speculate that fathers play a stronger role here because Orthodox Judaism demands more overt religious observance of men than of women; and so, children witness more religious behaviour from Jewish fathers than Jewish mothers. If so, then this is an instance of the CREDs effect at play (Section 5.2).

Recalling the discussion in Section 5.3 on parenting styles, recent research has also found different roles played by how fathers and mothers relate to their children. For example, although Bengtson et al. (Reference Bengtson, Putney and Harris2013) found that young adult children were no more likely to retain their mother’s religion than their father’s, a warm relationship with the father does seem to be more important than a warm relationship with the mother. That is, maternal warmth seems to make little difference to whether the child follows in their mother’s footsteps; but children are more likely to be religiously similar to their fathers if they enjoy a warm relationship with them. This is especially so for evangelical and mainline Protestants, and less so for Roman Catholics. The parental roles are reversed for Judaism and Mormonism, where what seems to matter is maternal warmth. These patterns are also likely subject to cultural differences. For example, Łysiak et al.’s (Reference Łysiak, Zarzycka and Puchalska-Wasyl2020) study on Polish adolescents that measured ‘real-time’ religious deconversionFootnote ³² – mainly away from Roman Catholicism – found maternal warmth to be protective against adolescent deconversion (though see Marek & Gurba, Reference Marek and Gurba2024 for evidence for the importance of closeness with the father).

Taken together, the evidence suggests that it is a mistake to focus solely on mothers. It is also important to note here that these comparisons between maternal and paternal roles risk obscuring the fact that, as discussed earlier, children are most likely to retain their parents’ religiosity when both parents agree with one another on religious matters and equally participate in the family’s religious life. Effective intergenerational transmission is not left to one or other parent, but is the shared responsibility of both. Indeed, other family members may also play a role; after all, vertical transmission includes not only that from parent to child, but also from grandparent to grandchild.

There is, perhaps understandably, not as much research on the role of grandparents; however, it seems likely that grandparents will become more important in rapidly secularising contexts, where older adults (i.e., of grandparental age) are much more likely to be religious than younger adults (i.e., of parental age). The LSOG is especially suited to examine grandparental effects, and Bengtson and colleagues have indeed found that grandparents – and especially grandmothers (Copen & Silverstein, Reference Copen and Silverstein2007) – exert an influence independent of parental influences. Furthermore, this is especially so in cases where the parents are divorced, implying that grandparents can mitigate the aforementioned negative effects of divorce on religiosity (Silverstein & Bengtson, Reference Silverstein, Bengtson, Alwin, Felmlee and Kreager2018a). As with parents, grandparental warmth with their grandchildren is important in enhancing their ability to positively affect their grandchildren’s religiosity. LSOG’s findings are echoed elsewhere too. For example, Gutierrez et al.’s (Reference Gutierrez, Goodwin, Kirkinis and Mattis2014) study of African American families found that grandmothers’ influence was comparable to fathers’ (both of which were dwarfed by mothers in this case; grandfathers had little effect). Legault-Leclair’s (Reference Legault-Leclair2025) Canadian study found a strong effect of having conversations about religion with grandparents in one’s teenage years predicted one’s religiosity; this effect was greater than that of having such conversations with fathers, and comparable to the effect of religious conversations with mothers – this study also found that conversation with siblings had a similar effect. All this suggests that grandparents may be helpful in parents’ efforts to pass down their faith.

6.1 Intergenerational Transmission in Minority Religious Communities

Most of the research on immigrant and minority religious communities has focussed on Muslims living in secular Western countries, and especially in Europe and North America. The basic observation is that intergenerational transmission of religion among Muslims is considerably higher than among Christians. Intergenerational retention of Islam remains high, not only in Muslim-majority countries, but also in Muslim-minority and even in rapidly secularising countries. According to 2018 British Social Attitudes data, for example, 93% of individuals who were brought up as Muslims still identify as Muslims, compared to 47% of those brought up Anglican (the established religion in England) and 55% of those brought up Roman Catholic. Only nonreligion was more successfully transmitted: 94% of those raised without any religion remain irreligious. Similar patterns can be seen across Europe (Spiegler et al., Reference Spiegler, Jonsson and Bracegirdle2025; Voas & Fleischmann, Reference Voas and Fleischmann2012) and also in the USA, though the difference between Islam and Christianity is less stark here (77% v. 73%; Lesage et al., Reference Lesage, Starr and Miner2025).

Several explanations have been offered for the intergenerational success of Islam in the West, some of which draw directly from the research literature we have already covered. The social importance of Islam has been emphasised, especially as religion and ethnicity are closely associated for Muslims, at least in the UK and Europe (e.g., Bangladeshi-Muslim and Pakistani-Muslim identities). This may account for the high levels of homophily among Muslims: Muslims are more likely to prefer to have Muslim friends than Christians are to prefer to have Christian ones (e.g., Leszczensky & Kretschmer, Reference Leszczensky and Kretschmer2022; Simsek et al., Reference Simsek, van Tubergen and Fleischmann2022). Although, as discussed in Section 4.2, peer conformity effects are relatively weak, certainly compared to parental effects, religious homophily for peers may have downstream effects on future marital and family life: religious homophily might lead to religious homogamy. Muslims have indeed been found to have stronger preferences to marry within the faith than Christians (e.g., Heyne et al., Reference Heyne, Kuhlemann and Kogan2025). Not only that, but Islam is exceptional in that men and women tend to be equally religious; and so, husbands and wives are likely to be similarly devout (Pew Research Center, 2016, March 22). As we have already noted earlier, marrying within the faith helps to maintain religiosity for the couple; religious homogamy also improves intergenerational religious transmission (see Section 5.5).

Besides the close social ties found in Muslim communities in the West, there are also some features inherent that might contribute to the high retention rate via the routes covered in previous sections. For example, the role that the Arabic language plays in Muslim life entails that additional religious education is often deemed necessary for Muslims to practice their faith at all, and parents feel obligated to make provisions for this additional education, whether at mosque-based madrassa/maktab, or in a more informal setting, such as a tutor’s home (Scourfield et al., Reference Scourfield, Gilliat-Ray, Khan and Otri2013). Quantitative data on the prevalence of such supplementary religious education is not yet available, but studies do find enrollment increasing in Europe and that many madrassas are oversubscribed (Alrasheedi & Almutawa, Reference Alrasheedi and Almutawa2020; Cherti & Bradley, Reference Cherti and Bradley2011). Besides the effects of religious education per se (Section 4.3), madrassas are also an additional source of religious socialisation for children, with adults (e.g., tutors) as well as peers (Sections 4.2 and 4.4). (We may note here that the Latter-day Saints, another minority religious community with traditionally high retention rates – though this is changing (Riess & Knoll, Reference Riess and Knollforthcoming) – also have a highly developed system of extra-curricular ‘seminary’ instruction for children and teens.)

Having mentioned the role of Arabic in Muslim prayer, it is also worth noting that the various behavioural aspects of Muslim religiosity are typically more involved than for Christians. The obligation to pray the salah five times daily, to fast during Ramadan, to maintain a halal diet are all inculcated from a young age, and maintained through adolescence. We have already seen evidence that childhood and adolescent religious involvement predicts adult religiosity (Section 4.4). Furthermore, as parents are also practising their faith, the effects of credibility-enhancing displays are also relevant here (Section 5.2).

So far, the situation for Muslims seems analogous to that of conservative Christians in the US, who also enjoy higher rates of intergenerational transmission than other Christians; Smith’s (Reference Smith2021) study of NSYR data has similarly found that this advantage comes largely from differences in religious socialisation. However, there may also be an effect of religious minoritisation per se, as indicated by the high intergenerational transmission rates among other religious minorities. For example, Scourfield et al. (Reference Scourfield, Taylor, Moore and Gilliat-Ray2012) found that in England and Wales, intergenerational transmission among other minority religions overall was close to 90%; similarly, Drouhot et al. (Reference Drouhot, Simon and Tiberj2023) found that intergenerational transmission of Judaism in France was at 84%. Molteni and van Tubergen’s (Reference Molteni and van Tubergen2022) study of Christian immigrants in thirty-three European countries may also be instructive here. They find that, unlike the case of Muslim immigrants, Christian immigrants do undergo significant intergenerational decline: however, this decline is significantly slower among Christian minorities, such as when Protestants move to Catholic countries. Scourfield et al. (Reference Scourfield, Gilliat-Ray, Khan and Otri2013) use the term ‘minority defence’ for the desire to and effort towards preserving cultural distinctiveness under pressure from dominant cultural influences. It is a phenomenon that applies not only to religion, but also to cultural traits more broadly: and analogous strategies are employed across these various cases, which mainly pertain to socialisation. Bunce and McElreath’s (Reference Bunce and McElreath2018) modeling work suggests that asymmetrical boundaries – through which minorities can cross into majority spaces, but not vice versa – are important for the maintenance of minority culture. For religious groups, this suggests a tension around exclusivity: on one hand, inclusivity might lead to growth by bringing in outsiders, but on the other hand, exclusivity might help to enhance intergenerational transmission within existing families.

6.2 Raising Highly Religious Youths

Given that highly religious children and adolescents are more likely to remain religious as adults than moderately or nominally religious children and adolescents (e.g., Lee et al., Reference 75Lee, Pearce and Schorpp2017; Watanabe & Todd, Reference Watanabe and Todd2025; see also Section 4.4), it may be important to understand what leads young people to be highly religious in the first place. By ‘highly religious’ we mean individuals for whom their religion is very important, evinced in their strength of religious belief and high levels of religious participation. Relatedly, highly religious groups are those that are strongly prescriptive in terms of belief and practice (e.g., Marks, Reference Marks2004). Besides intensifiers like ‘very’ and ‘highly’, researchers also write about ‘committed’ (Denton & Flory, Reference Denton and Flory2020), ‘devout’ (Smith et al., Reference Smith, Longest, Hill and Christofferson2014), ‘strict’ (de Bruin-Wassinkmaat et al., Reference de Bruin-Wassinkmaat, De Kock, Visser-Vogel, Bakker and Barnard2019), ‘conservative’ (Smith, Reference Smith2021), ‘traditional’ (Smith & Snell, Reference Smith and Snell2009), and ‘intense’ (Schnabel & Bock, Reference Schnabel and Bock2017) religiosity. These terms are related to one another and to our notion of ‘highly religious’, but they are not co-extensive with one another: they are sometimes used to include more specific religious positions, including literal approaches to understanding Scripture, positions of various moral and metaphysical issues, and even alignment with specific religious movements, such as evangelicalism – it is not our intention to conflate the highly religious with specific traditions.

There is, in a sense, nothing mysterious about highly religious youths. Typically, they come from highly religious families, who have successfully socialised them in the ways described in Sections 4 and 5. However, it is instructive to begin with highly committed adults, and trace back to how they became so committed – or, more likely, how they remained so committed. Denton and Flory’s (Reference Denton and Flory2020) study on four waves on NSYR data is helpful here. By Wave 4, the participants – who started out a decade back, aged 13 to 17 years – are now in their mid-twenties. 29% are characterised as having had ‘high stable’ religious trajectories, having started as the most religious adolescents (i.e., attend services weekly, regards faith as very important, prays daily) and managed to maintain high levels of religiosity as emerging adults. Unsurprisingly, some of the factors we have already considered in Sections 4 and 5 emerge here as distinguishing characteristics of individuals on this religious trajectory compared to those in other religious trajectories (e.g., starting out moderately highly religious but then experiences steep religious decline; more gradual religious decline). Those who remained highly committed did have parents who attended religious services frequently, including with them; but the parents of those who experienced steep or gradual decline were not much less religious. The factor that seemed to make the most difference is whether or not, and how frequently, religious and spiritual matters were discussed at home during adolescence (see Section 5.2). Just over 60% of highly and stably committed individuals came from such households, in comparison to ~35% of those who experienced religious decline. Talking about religion at home – God, prayer, Scriptures, and so on – is not a guarantee that adolescents will retain their faith in adulthood, but it is a strong contributing factor. As we have already seen in Section 5.3, religious conversations in the homes of highly religious youth also take on a particular form, which Dollahite and Thatcher (Reference Dollahite and Thatcher2008) describe as ‘youth-focussed’ rather than ‘parent-focussed’. Recall that youth-focussed conversations are open conversations, instigated and directed, at least in part, by the adolescents, who talk and not just listen to what the parents say; the topics of conversation are also related to their real lives, and the conversation serves to nurture the relationship between parent and child.

6.3 The Intergenerational Transmission of Nonreligion

For most of this Element, we have assumed that the religiosity to be transmitted comprises what we might call ‘positive’ religious beliefs, behaviours, and identities. We do not mean this morally, or otherwise normatively. Atheism, being at least the absence of theistic belief, and perhaps also its overt rejection, counts here as a negative religious belief.Footnote ³³ Similarly, identifying as nonreligious in various ways – whether as atheists, agnostics, or simply ‘no religion’ – is also negative. Sometimes, and perhaps increasingly, the religiosity to be transmitted is negative in this way.

Given the sociological trends, it would seem that the intergenerational transmission of nonreligion is easy. In contrast to the intergenerational religious decline that we see in sociological data, nonreligion seems ‘sticky’ (Strhan & Shillitoe, Reference Strhan and Shillitoe2019). Pew Research Center’s 2023/2024 study found that this pattern held in most of the world: in the fifteen qualifying countries,Footnote ³⁴ intergenerational retention of nonreligion in ten of the countries was about 80%, and the lowest retention rate, in Argentina, was 63% (Lesage et al., Reference Lesage, Starr and Miner2025, March 26). This study asked participants if they were raised in a particular religion, which is different from asking about parental religion. Even with the stickiness of nonreligion, we might find an effect of homogamy here. Pew’s UK estimate for the nonreligious retention rate is 87%, but the 2018 British Social Attitudes data has it at 94% if both parents are nonreligious (Voas & Bruce, Reference 84Voas, Bruce, Curtis, Clery, Perry, Phillips and Rahim2019). Similarly, Pew has the nonreligion retention rate in the US at 73% and Canada 78%; but the 2019 Millennial Trends Survey has it at 91% and 89%, respectively, when both parents are nonreligious (Wilkins-Laflamme, Reference Wilkins-Laflamme2020). Wilkins-Laflamme (Reference Wilkins-Laflamme2020) overtly notes that nonreligious retention is higher when both parents are nonreligious than when just one is. McPhail’s (Reference McPhail2019) analysis of 2008 ISSP data corroborates this more broadly. In other words, religious homogamy cuts both ways: it applies to nonreligion as well as to religion (see Section 5.5). LSOG data suggests that, at least in the USA, fathers’ nonreligion is more influential than mothers’ nonreligion (Bengtson & Silverstein, Reference Bengtson, Silverstein and Timonen2018). The LSOG findings on the importance of parental closeness in the intergenerational transmission of religion (Section 5.4) also apply to nonreligion.

7.1 Practical Conclusions for Parents and Religious Leaders

Although this Element is intended primarily for fellow social and psychological scientists, we expect that there will also be parents and religious leaders among our readers, who might be looking for practical advice about how to pass down the faith to the next generation. A few things seem clear for us, from the research literature:

1. 1. The sociological trends – globally – are towards secularisation over time, as societies grow wealthier, more secure, and better educated. And so, the challenge of transmitting religiosity from generation to generation is a difficult one, which may get more difficult, even among religious groups that currently still enjoy a high rate of intergenerational religious transmission.Footnote ³⁵ If parents and religious leaders want their children to retain the religiosity of their upbringing, they need to work at it. Nevertheless, there is no silver bullet, no ‘one weird trick’, no magic formula that guarantees successful religious transmission. Lots of parents who do everything right to ‘increase the odds’ will end up with children, and moreover grandchildren, who are either less religious than they had hoped, or not religious at all.

2. 2. High religious commitment has a protective effect against decline. Those for whom religious belief, behaviour, and socialisation were very important parts of their lives as children are more likely to remain religious, even highly religious, as adults. Those who remain religious tend to have grown up with very strong religious identities, frequent religious service attendance, participation in youth groups, and socialise with other religious people. There is a substantial difference between highly religious and even moderately religious families.

3. 3. Parents have a very strong role to play in what they do, what they say, and what they decide on behalf of their children. The active religious participation of parents is important: parents who successfully transmit their religiosity to their children are those who show their children that religion is important. Talking about religion is important too: children who retain their childhood religiosity tend to come from families where religion is discussed frequently at home. It is also important to attend to how these conversations go. Parents often subtly express uncertainty on religious matters, and this leads to weaker belief among even young children. Parents can also be too didactic and one-sided in their talking to adolescents about religion: instead, children should be active participants in and instigators of discussions about religion, and these conversations should relate religion to children’s daily lives. Indeed, parenting style – and especially, authoritative parenting-style – is helpful for passing down one’s religion (and nonreligion). Finally, permissiveness about children’s and adolescents’ religious participation predicts decline. Children whose parents decide to prioritise and support their children’s religious participation are more likely to retain their childhood religion.

4. 4. Religious communities – churches, mosques, synagogues, temples, and so forth – do have a role to play, but this role is relatively minor, compared to the role of parents. Participation in youth groups – and having peers who do too – does seem to predict religious retention; religious groups that want to prioritise intergenerational transmissions should therefore make provisions for such groups. However, given the importance of parents in religious socialisation, the main role of religious leaders and communities should be to support parents in their efforts.

7.2 Risk Factors for Religious Disaffiliation

Given secularising trends, it cannot be taken for granted that children who are brought up religious will remain religious (see Point 1 in 7.1 above), especially if they are brought up in a majority religion in the West. However, there are also some more specific signs of future religious decline, and some sociodemographic subgroups are particularly susceptible to leaving their childhood religion. For example, certainly in the USA, White males are most susceptible to losing their religion (Hardy & Taylor, Reference Hardy and Taylor2024). LGBTQ+ individuals are also highly likely to religiously disaffiliate (Woodell & Schwadell, Reference 85Woodell and Schwadel2020), which is perhaps unsurprising given that sexual minorities are often stigmatised or marginalised in religious communities (Lefevor et al., Reference Lefevor, Bouton, Davis, Skidmore and Meyer2025).

As we have already considered in Section 5.5, certain family structures are risk factors of religious disaffiliation. Children who grow up in married two-parent households are most likely to retain their religiosity; and divorce is a risk factor for religious disaffiliation. Similarly, religiously mixed marriages are challenging contexts for transmitting religiosity to the next generation, especially if the marriage is between a religious person and a nonreligious one. Parents in mixed marriages who nevertheless want to transmit religiosity will, therefore, need to find some means to compensate for this – as we note in Section 5.6, grandparents may be helpful here.

Research on the processes of deconversion or de-identification, particularly in adolescence and emerging adulthood, also suggests some individual-level risk factors. Overall, the lesson is that religious de-identification is rarely sudden, but is instead preceded by declines in various aspects of religiosity. For example, in their study of four waves of NSYR data, Hardy et al. (Reference Hardy, Hendricks, Nelson and Schwadel2023) found that declines in religious service attendance, prayer, and scripture study in adolescence all foreshadow de-identification by young adulthood (see also Watanabe & Todd, Reference Watanabe and Todd2025). Looking at national-level data from the World Values Survey, Stolz et al. (Reference Stolz, de Graaf, Hackett and Antonietti2025) similarly finds that declines in public religious participation precedes declines in religious identification.

Hardy et al.’s (Reference Hardy, Hendricks, Nelson and Schwadel2023) study also provides some evidence that increases in doubt during adolescence predicts de-identification later on; however, being highly doubtful early on is itself predictive of later de-identification, regardless of trajectory. Watanabe and Todd’s (Reference Watanabe and Todd2025) analyses of the same dataset suggest that maintaining or even increasing religious engagement – such as religious service attendance – offers some protection when there is growing doubt, but doubt remains a powerful predictor of religious de-identification.

There has also been some research on the reasons people give for leaving the religion of their childhood. When asked why they left, Van Tongeren et al.’s (Reference Van Tongeren, Brady, Van Duinen, Burke and Ravey2025; see also McLaughlin et al., Reference McLaughlin, Van Tongeren, Teahan, Davis, Rice and DeWall2022) participants from the USA and UK tended to talk about religious beliefs losing their force, whether in intellectual terms or in terms of consistency with their other beliefs and values. They draw a distinction between ‘belief misalignment’ and ‘disillusionment’, with the latter being associated with more generally ‘outgrowing’ their childhood religion, but these seem to overlap (e.g., talk of childhood religion failing to account for how they see the world), and McLaughlin et al. (Reference McLaughlin, Van Tongeren, Teahan, Davis, Rice and DeWall2022) do elide the two.

Other commonly cited reasons were disappointments – especially moral disappointments – with religious individuals, leaders, and organisations; and also religious individuals’ and institutions’ association with suffering, whether as perpetrators (e.g., in cases of clergy abuse, experienced or observed) or as poor pastoral responders (e.g., when the participants themselves were suffering). In other words, people who have left religion tend to cite, as reasons, struggles with belief as well as with other religious people’s moral failures. This latter phenomenon has been called ‘credibility undermining displays’ (CRUDs; e.g., Turpin & Willard, Reference Turpin and Willard2022), which is the inverse of credibility-enhancing displays (see Section 5.2). Research tends to focus on extreme examples, such as abuse by clergy, but it applies to religious hypocrisy more broadly. For those interested in reducing religious de-identification – and thus bolstering intergenerational religious transmission – these themes may therefore be important to discuss, whether at home or in religious services, whether by parents or by religious leaders.

7.3 Future Directions

In Section 1, we mentioned that this Element is aimed mainly at researchers, and to encourage more research on the intergenerational transmission of religion. Throughout the Element, we have noted gaps in the research literature, which present obvious opportunities for further research. In this final section, we bring some of these together.

Understanding the intergenerational transmission of religion clearly requires longitudinal data, not just from large-scale sociological surveys but also from smaller but more detailed developmental psychological studies. There are still not very many of either, and these mainly focus on Western European and North American populations. Longitudinal research is difficult and expensive, but it is essential, and there needs to be more of it.

Although we know quite a lot about who the most important sources of religious socialisation are – namely, parents and peers – we lack textured detail about how their influence works in different contexts. For example, in Section 4, we saw that there are peer conformity effects in play, and also that youth group involvement is an important predictor of religious retention, but we are only beginning to learn about the kinds of interactions that adolescents have with one another that best encourage religiosity. Similarly, Section 5 establishes the critical importance of parents, including especially of religious discussion in the home, but there is very little research on the nature of these conversations, and what work there is risks being outdated now. As parents of young children, we know that children ask all kinds of interesting – and difficult! – questions about religion: what we don’t know much about is how these conversations ought to proceed, if we want to encourage our children’s religiosity.

There is also not very much detail about what else happens in the home that encourages religious retention. At best, studies ask whether families pray together, such as during meals. But mealtime prayers represent a single example of family-based religious rituals and activities. There is also some qualitative work on religion at home, but this quantitative research remains under-developed. More research is required to capture the diversity of such activities – from the reading of religious books to prayerful bedtime routines – and their effects on intergenerational religious transmission.

Although much of the longitudinal evidence suggests that parents play the strongest role, it is also true that people who have left the religions of their upbringing often cite reasons for their leaving that have to do with the religious communities, including feeling out of place or repressed by them, or disagreeing with various moral and doctrinal teachings (e.g., Hardy & Longo, Reference Hardy, Longo, Hardy and Longo2021; Packard & Fergusson, Reference Packard and Ferguson2019). While it is possible that such retrospective and introspective self-reports are unreliable, it is also worth exploring these factors in future longitudinal studies.

It seems likely that social media will become more important sources of religious socialisation, not only from known peers, but also through content produced by other individuals and organisations. It is important not only to understand whether and how social media influences children, adolescents, and emerging adults, but also how parents and other caregivers might best respond to undesirable influences.

As religious landscapes shift towards increasing secularism and pluralism, research on religious minorities will become more important, especially because religious groups that were historically in the majority are increasingly being themselves minority. Research on nonreligious transmission and socialisation has begun, but this is mainly focussed on the already secularised West: as secularisation increases in other parts of the world, it is also important to study nonreligious socialisation in those contexts, especially where there might be hostility against nonreligion.

These are just a few of the topics that we hope to see developed in the research literature in the coming years. Perhaps this Element might inspire more research in other directions, which we have not yet imagined.