A systematic review of the association between religiousness and children’s prosociality

ABSTRACT

Religiousness can guide children’s morality and behaviors. To better understand how the growing research on religion relates to child development, and in particular, prosociality, a systematic review was conducted. Three research questions guided the systematic review. First, how are religiousness and children’s prosociality measured? Second, what is the association between religiousness and prosociality? Finally, is there variation in that association based on moderators or study design? Eligible studies included published research on the association between religiousness and children’s prosociality (n = 135), of which nine were included in the final review based on inclusion/exclusion criteria. Narrative synthesis found that religiousness is mostly measured through religious affiliation, and all studies measured prosociality through general prosocial behavior, while the others also focused on empathy and/or altruism. Five of the nine studies found no association between religiousness and children’s prosociality. Variation across overall study quality, demographic variables, and methodological characteristics is explored. The review concludes with implications for future research, such as a focus on the early years and the targets of children’s prosociality.

KEYWORDS:

• Religiousness
• prosociality
• systematic review
• narrative synthesis
• children

Introduction

Sharing beliefs through religion makes people part of a society, part of a group (Bar-Tal, 2000; Ysseldyk et al., 2010). This review seeks to find out how religiousness relates to children’s prosociality, and in particular, who are the targets of these prosocial acts. Does being part of a religious group lead children to be more prosocial, and if so, are they more likely to be prosocial to ingroup as well as outgroup members?

Developing prosociality

Prosocial behaviors are cooperative, or generous behaviors or attitudes (Shariff et al., 2015) where behavior benefits another (Eisenberg et al., 2015). This includes sharing, giving up your time to help, cooperating, responding to others’ needs, and refraining from lying and cheating. Through prosociality, children show concern for others, such as showing physical or verbal comfort, helping, indirect helping, sharing, distraction from pain, protection or defending. Prosocial behavior is distinct from altruism, and empathy (Bierhoff, 2002), although some studies use these terms interchangeably. Prosociality can be distinguished from empathic behavior since empathic behavior involves an emotional arousal seen through facial or vocal expression that reflects sympathy for the victim (Zahn-Waxler et al., 1992). Altruism has been considered a subset of prosociality since prosocial behavior can be due to egoistic motivations, whereas altruistic behavior benefits the other as an end in itself (Batson et al., 2007). Egoistic motivations can be related to material benefits, social or self-rewards, avoidance of punishments or reduction of aversive-arousal (Batson et al., 2007). Unless behavior motivation is assessed, it is hard to establish whether altruism or prosociality is measured.

Although prosociality increases with age, there is evidence of prosocial understanding as early as the first year of life. For children aged 2–3 years, there is an increase in empathic behavior directed mainly towards the mother (Zahn-Waxler et al., 1992). It is only by 3 years that prosocial behaviors are directed towards a stranger (Robinson et al., 2001). Although prosociality, such as helping and cooperating, increases with age (Köster et al., 2016), as children grow up, these prosocial behaviors will be discriminated depending on whether the receiver is an ingroup member or a stranger (Hay & Cook, 2007). The Developmental Peace Building Model (DPM) illustrates how children develop prosociality through interpersonal relationships (Taylor, 2020). These relationships help children effect structural and cultural change in the exosystem (e.g., teacher’s home life and activities of the school board which did not contain the child but whose events effect the child) and macrosystem (that is, the social organization, belief system and lifestyle of the society in which the child lives, Bronfenbrenner, 2005). Children are not just influenced by external factors, but the DPM construes children as active participants that can promote social change through prosocial behaviors. The DPM considers the developmental process and calls for the need to focus on the “targets” of prosocial behavior to predict how social cohesion can be promoted when there is intergroup conflict. To provide a comprehensive synthesis of research on children’s prosociality, the targets of prosocial behavior need to be underscored.

Religious development

The child develops religious constructs within the community through shared practices, rituals, stories and symbols (Boyatzis, 2005). This development is framed through cognitive-developmental explanations and social influence. Studies of children’s drawings of God and the house of God (Harms, 1944; Konyushkova et al., 2016; Pnevmatikos, 2002), and on the development of the prayer concept (Elkind, 1964; Long et al., 1967; Spilka & Ladd, 2013) show children’s stage-like religious development from concrete egocentric to abstract sociocentric notions, e.g., from God as a human living next door, to God with supernatural powers living in the clouds. Research also converges on the generational transmission of religiosity (Boyatzis & Janicki, 2003). For instance, parental religious importance, church attendance (Devine, 2013), observation of parents fasting and praying (Henrich, 2009; Kelley et al., 2021; Lanman & Buhrmester, 2016) imply that parents enhance religious transmission. It is noteworthy to synthesize how religiousness in children is measured or if this is studied through parental religiosity.

Religiousness and prosociality

Religiousness is related to increased prosocial behaviors. Since most religious congregations organize or sponsor community services, for instance, fund-raising, volunteering, or contributing to a soup-kitchen, shows that religious affiliations may increase prosociality of their members. Given that prosociality is a behavior that benefits another (Eisenberg et al., 2015), these community services promoted by religious congregations are encouraging prosociality (McClure, 2017). Religiousness increased prosociality in emerging adults (Norenzayan & Shariff, 2008), clergy students (Ahmed, 2009), and the general adult population (Shariff et al., 2015).

Manifestations of religiousness by all religions in the world have been attributed to five dimensions: ideological, ritualistic, experiential, intellectual, and consequential (Glock, 1962; Stark & Glock, 1968). These aspects of religiousness increase motivation for prosociality at the cost of self-interests (Norenzayan & Shariff, 2008; Ruffle & Sosis, 2007; Xygalatas et al., 2013). Moreover, although in the last 15 years, there has been a growing interest in the link between religion and prosociality, few studies focused on children (Galen, 2012; Oviedo, 2016). There is a robust effect of religious priming on prosociality for religious adults (Shariff et al., 2015), however, this meta-analysis and a systematic review on religious attitudes and prosocial behavior (Oviedo, 2016) did not include studies with child participants. This raises important questions about the development of the association between religiousness and children’s prosociality.

Current study

A systematic review is a structured and clear methodology for selecting studies to minimize bias which leads to more reliable findings from which inferences can be made (Higgins et al., 2019). The main aims of this systematic review were to:

1. Identify how religiousness and prosociality are measured. Analyzing how these constructs are measured will shed light on the caveats and aptness of the association.

2. Examine the direction of the association of religiousness and prosociality to find out possible conditions for this association.

3. Explore potential moderators and variability across studies (e.g., quality, study demographics—age and gender, and design) to ensure the reliability and validity of the direction given in the studies.

Method

The protocol for this systematic review was registered with PROSPERO and published on OSF¹ on 5th May 2021. This review followed the PRISMA guidelines for systematic reviews (Moher et al., 2014; Page et al., 2021).

Search strategy

We conducted comprehensive searches of the ERIC, Web of Science, PsychINFO, MEDLINE, EMBASE, ASSIA, Society for Research in Child Development (SRCD) through Wiley Online, OATD, DOAJ, Dissertation Proquest, Dissertation and Theses at UCD, and OpenGrey Databases. Grey literature was searched for through Dissertations ProQuest, Conference abstracts, book chapters publications “in press,” and reports. Thus, non-peer-reviewed journals and other unpublished papers were included to avoid publication bias. We did not impose restrictions on the search date or language. The publication period was prior to the end of the search on 14th May 2021.

Across titles and abstracts, the primary search terms included three main constructs and their variations: Children (child*, early childhood, preschool, student, pupil, kindergarten); religious (relig*, Christian, Orthodox, Protestant, Muslim, Hindu*, Jew*, Buddh*, Sikh*); and prosocial behavior (prosocial behav*, altruism, kind, sharing, charitable). Non-child samples were excluded through the NOT boolean (College, university, Third level, adol*, youth, polytechnic, adults, higher education, young people, young person). The screening, the full-text review, the data extraction, and the quality assessment was conducted through Covidence extraction 2.0.

Inclusion and exclusion criteria

Studies were included if they met the following criteria:

1. The participants were children aged 3–10 years or adult reporting about children in this age group. Studies that did not specify the ages of their samples were excluded if the age could not be reasonably inferred (e.g., described as “first graders” in the US system, which is typically aged 5–6). An inclusive approach was used for this review. For instance, studies like Pandya (2017), where the youngest participants were aged 10 and at the borderline age of this review, were included in this systematic review.

2. For the religiousness measure, studies that measured religiousness as a parental construct, e.g., parental religious affiliation or religious attendance, were included in the synthesis, given the empirical agreement on the generational transmission of religiosity.

3. They entailed empirical research.

4. They analyzed the relationship between religiousness as a predictor or covariate and at least one quantitative variable measuring prosociality. Studies that examine children’s prosociality, but do not relate it to religiousness, were excluded.

5. They examined the target of prosocial behaviors. Studies were excluded if they did not specifically examine the relationships between religiousness and children’s prosociality.

In the case of a retracted study, the decision to exclude a study was taken based on the reason for the retraction (Faggion, 2019). Figure 1 presents a flowchart of the eligibility process for this review. Testing inter-rater reliability at different time points is methodologically sound to ensure consistency and clarity at different stages of the systematic review (Belur et al., 2021). Therefore, there were three phases of inter-rater reliability of title and abstract screening. First, two blinded reviewers conducted the title and abstract screening of 20% of the studies, resulting in a Cohen’s Kappa of .53, indicating a “weak” level of agreement (McHugh, 2012). Seven conflicts were discussed between the two reviewers, e.g., that religiousness could also have been measured as a covariate, and the inclusion/exclusion criteria were consequently revised. Second, the reviewers screened another 20% of the studies, with a Cohen’s Kappa of .63 which is a “moderate” level of agreement (McHugh, 2012). Inclusion and exclusion criteria were again refined, e.g., dealing with unknown children’s age. Finally, using the refined set of inclusion/exclusion criteria, the two reviewers independently screened the remaining studies (Cohen’s Kappa was .92). Of the 135 identified studies (after duplicates were removed), 113 publications were excluded at the title/abstract screening stage, leaving 22 studies for the full-text review stage.

Figure 1. PRISMA flow diagram showing the selection of studies for the review.

For the full-text screening, a sample of full-text studies was independently screened by two reviewers until concordance was achieved (∼25%; 5/13 full-text studies).² After resolving disagreement through discussion and obtaining the Cohen’s Kappa of 1.00, 13 articles were excluded leaving 9 articles eligible for data extraction (Page et al., 2021).

Data extraction

Data from 9 articles were extracted by the two independent reviewers through Covidence. The data extracted was coded in four sections: (1) general information data, (2) population and setting, (3) methods and measures, and (4) results. Data for Aim 1 were primarily extracted from the method section, and data for Aim 2 from the results section, and for Aim 3 from all sections.

Study quality

Consistent with best practice, data extraction was conducted prior to rating the study quality (Higgins et al., 2019). The quality assessment for each study analyzed selection bias, ethical issues, and soundness of the interpretive approach (Whiting et al., 2011). Studies were classified as either low risk, high risk, or unsure, for each of these dimensions by the first reviewer, with 25% by the second reviewer. Two inconsistencies in the quality assessment were resolved through a consensus meeting between the reviewers. One of the studies extracted was a retracted study (Decety et al., 2015). The Cochrane Handbook (Higgins et al., 2019) explains the reasons for retracted studies but offers no clear guidelines on how to proceed. Thus, the reason for the retraction was taken into consideration. The study was not retracted due to the measures used but because of inaccurate data analysis. Since Decety et al. (2015) reported the measures of religiousness and prosociality and Shariff et al. (2016) reported the reanalysis results, it was decided that this study will be referred to as Decety et al. (2015)/Shariff et al. (2016). This systematic review reports the measures used in Decety et al.’s study but the results of the association as reanalyzed by Shariff et al.

Data synthesis

Narrative synthesis was deemed suited for this review since the authors were interested in studying relationships within and between studies through an examination of moderator variables (Popay et al., 2006). This approach adds meaning to the findings that can be used to plan further research (Roger et al., 2009).

Results

Since empirical research on the association of religiousness to prosociality was reviewed in a systematic way, the results of the studies can be compared (refer to Table A1 in Appendix).

Study characteristics

This systematic review includes 9 studies published between 1976 and 2021 across 19 countries (Australia, Canada, China, Denmark, Egypt, France, Germany, India, Israel, Japan, Jordan, Northern Ireland, Saudi Arabia, Singapore, South Africa, Turkey, Sweden, United Kingdom, and United States), and 8 religious affiliations defined by the authors of the studies (Buddhism, Protestantism, Christianity, Evangelicalism, Judaism, Catholicism, Islam, Hinduism, spirituality, not religious or no religious affiliation). Studies in this review consisted of mostly WEIRD samples, with only one study of completely non-WEIRD samples and two studies with mixed samples.

Aim 1

First, we identified how religiousness and children’s prosociality were measured.

Measures of religiousness

Six dimensions listed below measured religiousness across the included studies (Table 1). Six of the nine studies assessed more than one dimension of religiousness. Figure 2 indicates which of these dimensions were measured in the studies and if these were measured through parental or children’s reporting. This shows whether children’s or parents’ religiosity was measured.

Figure 2. Figure showing the six different dimensions used to measure religiousness and the informer of the dimension. Parental report indicates that parental religiousness was measured. *The child report was administered by an adult. Italics indicate the studies that showed no association.

Table 1. Dimensions of religiousness measure.

Note: The shaded cells indicate the use of the dimension.

Religious affiliation

This variable refers to the religious affiliation or denomination of the child and their family, e.g., Muslim, Catholic, Protestant, etc. For three studies, this was the only religiousness measure but for the other studies, affiliation was used in conjunction with the other measures mentioned below. As can be seen in Figure 2, in most studies, religious affiliation was measured through the type of school attended by the participants. Only in one study (Pandya, 2017), where the youngest participants are 10 years old, were the children asked for their religious affiliation.

Religious attendance

This is a count of attendance to a religious practice reported by adults for the three studies that operationalized this dimension. Religious attendance was measured through the Duke University Religious Index (DUREL, Koenig & Büssing, 2010) in a questionnaire answered by the parents through the item “How often do you attend church or other religious meetings?” Religious attendance was also measured and labeled as part of the religious attitude towards religion and values construct with the item “I attend religious services two or three times a week” (Decety et al., 2015/Shariff et al., 2016; Kingston & Medlin, 2005). In one study, religious attendance consisted of a spiritual, educational program where attendance was reported by the spiritual trainer (Pandya, 2017).

Private religious practices

This category refers to the routines, such as praying, practiced by the family of the children. These practices might occur inside or outside the place of worship. They are also referred to as family religious practices because they specifically measure practices at home, such as reading the bible as a family (Crosby & Smith, 2015). The DUREL (Koenig & Büssing, 2010) was used to measure private practice through the item “How often do you spend time in private religious activities, such as prayer, meditation, or Bible study?” (Decety et al., 2015/Shariff et al., 2016; Pandya, 2017). The child’s parents/guardians also kept a self-practice record card where they reported the regularity of the child’s practices of the homework exercises given as part of the program (Pandya, 2017).

Religious beliefs

The online Oxford Languages dictionary defines belief as “an acceptance that something exists or is true, especially one without proof” and “trust, faith, or confidence in (someone or something).” Items used to measure religious belief were: “Saying my prayers helps me a lot” and “God is very real to me” (Crosby & Smith, 2015). In this study, the construct of spirituality was operationalized through items such as “I feel like God is close to me” or “Saying my prayers helps me a lot” which are also measuring religious beliefs. The construct of religious attitudes and values that measured religious belief was operationalized through eight items: e.g., “My religious faith is important to me” and measured on a five-point scale (Kingston & Medlin, 2005). Religious beliefs were also measured through an item on the DUREL: “My religious beliefs are what really lie behind my whole approach to life” (Decety et al., 2015/Shariff et al., 2016; Pandya, 2017).

Religious experience/spirituality

This dimension measures the experience of the divine/higher power in life or interconnection with nature in the case of spirituality—the feelings of awe and gratitude to a higher power (Büssing et al., 2018). Two studies label this construct as spirituality, although it is measured in different ways. Spirituality was measured through the 16-item Spirituality Assessment Scale (Liu & Robertson, 2011) that measures interconnection with human beings, with nature and with a higher power on a five-point scale (Pandya, 2017). The construct of spirituality was also used to measure religious experience and religious beliefs (Crosby & Smith, 2015). The three items that refer to religious experience are “I talk to God a lot,” “God talks to me,” and “I feel like God is close to me.” The DUREL also measures religious experience through the items “In my life, I experience the presence of the Divine” and “I try hard to carry my religion over into all other dealings in life” (Decety et al., 2015/Shariff et al., 2016; Pandya, 2017).

Consequences of religiousness

This measure determines the sense of belongingness participants and their family feel towards their congregation and/or members of other religious affiliations. This construct labeled as “church support” was measured through items such as “My church is like a family,” operationalized as the expression of belongingness and concern for other churchgoers (Crosby & Smith, 2015). Other studies measured essentialist beliefs of the children towards members of another religious affiliation (O’Driscoll et al., 2021) or ingroup symbol preference (Shamoa-Nir et al., 2020).

These results show that for all the dimensions, expect for the Consequence of Religiousness, some studies measured parental religiousness. Only four studies measured children’s religiousness across the six dimensions. Children’s religiousness was measured through child reporting because either the children were 10 years and over (Pandya, 2017) or an adult administered the questionnaire to the children (Crosby & Smith, 2015; O’Driscoll et al., 2021). Another study measured children’s religiousness through a child task (Shamoa-Nir et al., 2020). Thus, parental religiousness is mostly measured through the studies.

Measures of prosociality

Three subsets of prosociality (i.e., general prosocial behavior, altruism, and empathy) were assessed across four types of measurement (i.e., parent report, child report, quasi-experimental, and neurophysiological). When the participants were 9 years or less, the child report was administered by an adult. Otherwise, quasi-experiments or parental reports measured children’s prosociality.

General prosocial behavior

This subset refers to measures that grouped different types of behaviors such as sharing, helping, showing kindness, and being empathic, into one general construct. All studies measured prosociality as a general construct, while a few studies included empathy and/or altruism too. The Strengths and Difficulties questionnaire (Goodman, 1997 in Crosby & Smith, 2015) and Prosocial Personality Battery (Penner et al., 1995 in Pandya, 2017) were used as self-report measures. Parental reports on children’s prosociality used a 5-point scale (Kingston & Medlin, 2005). Prosociality or prosocial giving was mostly measured through a quasi-experiment. The dictator game was used in three studies (Decety et al., 2015/Shariff et al., 2016; Robbins et al., 2016; Sabato & Kogut, 2018). In the other four studies, children had to distribute stickers, candy, or peanuts to fictitious others. One study used helping as a prosocial measure (although the study labeled it as an altruistic construct) where the children were asked to help the experimenter stack slides into different piles (Panofsky, 1976).

Altruism

This variable measures how much children can help others and prioritize them over their needs. Different child report measures were used: the Child Altruism inventory (Ma & Leung, 1991 in Crosby & Smith, 2015) and Self-report Altruism Scale (Rushton et al., 1981). Children rated their use of altruistic behavior/tendency on a five-point scale (Pandya, 2017). The Children’s altruism measure (Litvack-Miller, 1991) required children to decide whether they will be helping a person, ignoring the request for help or get help after reading a story about their situation (Kingston & Medlin, 2005). Altruism was also measured through a helping task (adapted from Rubin & Schneider, 1973), where participants were asked reasons for sharing and helping the subject (Panofsky, 1976). Unless the measure used was an established altruism scale as the ones referred to above, or a measure for the behavior motivation, it could not be established as a construct of altruism, and it was classified as a general prosocial construct.

Empathy

This subset measures how much children understand and respond to other’s emotions. Parental reports measured empathy in one study (Decety et al., 2015/Shariff et al., 2016): the questionnaire of cognitive and affective empathy (Reniers et al., 2011) assessed children’s empathy, and the justice sensitivity inventory (Schmitt et al., 2005) measured children’s sensitivity to injustice. Children's empathy was measured through the Adapted interpersonal reactivity index (Davis, 1980, 1983), adapted for use with children (Kingston & Medlin, 2005; Litvack-Miller, 1991). This self-report measure was based on the child’s perspective taking, empathic concern, personal distress, and emotions expressed during fantasy-related activities such as watching a movie. Another self-report measured outgroup-directed empathy (O’Driscoll et al., 2021; Turner et al., 2013). In an empathy task (Feshbach & Roe, 1968), the children heard a narration about the interest of a child and had to match the feelings of the subject (Panofsky, 1976). Empathy was also constructed through neurophysiological measures where the children were shown scenarios depicting harm to people and objects, and activity in the amygdala, insula, and temporal poles were detected using the moral sensitivity task (Decety et al., 2015/Shariff et al., 2016).

These measures show that prosociality, whether determined through a general construct of prosocial behavior or also through altruism and/or empathy, data was mostly collected through the child. In fact, none of the studies relied solely on parental data as a measure of prosociality (Figure 3).

Figure 3. Figure showing the three subsets measuring prosociality. The parental report measures the children’s prosocial behaviors. *The child report was administered by an adult. Italics indicate the studies that showed no association.

Aim 2

Second, we examined the association of religiousness to children’s prosociality (refer to Table 2). Four studies found a positive association, while five studies found no association between religiousness and children’s prosociality. One particular study originally found a negative association between religiousness and children’s altruism (Decety et al., 2015). Upon reanalysis by Shariff et al. (2016), a retraction notice was issued admitting incorrect analysis (Decety et al., 2019). Shariff and colleagues found no effect of religiousness on prosociality across all the six countries assessed. The reanalysis found a non-significant negative effect so that a small increase in religiosity (SD = 1) predicted the sharing of 0.2 stickers. A study which reported no association found that it was not religious affiliation that affected prosociality, but the different types of prosocial behaviors that were measured (Panofsky, 1976). Children from one religious affiliation did not behave any different than children from other cultures in respect to their prosociality, so this study reported no association between religiousness and prosociality (Robbins et al., 2016). In two studies, the direct effect of religiousness to prosocial behaviors was not significant, however the models produced in the studies elaborate more on the conditions of this association (O’Driscoll et al., 2021; Shamoa-Nir et al., 2020). Heterogeneity across studies also suggest that there may be variability related to study quality, demographic variables, methodology, and variables measured. Given that less than half of the studies found a positive association, we cannot conclude that there is an association between religiousness and children’s prosociality.

Table 2. Table of the association of religiousness to children’s prosociality.

Association	Frequency of studies	Studies; APA citation
No association	5	Decety et al. (2015)/Shariff et al. (2016); Panofsky (1976); Robbins et al. (2016); Shamoa-Nir et al. (2020); O’Driscoll et al. (2021)
Negative association	0
Positive association	4	Crosby and Smith (2015); Kingston and Medlin (2005); Pandya (2017); Sabato and Kogut (2018)
Total	9

Aim 3

Third, we explored potential moderators and variability across studies (e.g., quality, participant demographics—age and gender, and study design).

Study quality

The majority of studies included in the review were of good quality; that is, rated as “low risk” across the three quality dimensions: selection bias, ethical issues and interpretative approach. For selection bias, unclear risk arose around the use of convenience sampling (Kingston & Medlin, 2005), or relatively small samples (e.g., n = 29; Robbins et al., 2016). For ethical issues, unclear risk arose based on the lack of explicitly noting of ethical approval, although it might be assumed to have been secured such as part of a doctoral dissertation (Panofsky, 1976). For interpretive approach, one study was rated as “unclear risk” (Decety et al., 2015); that is, the quality on this dimension was a concern. In fact, this study was retracted by the journal after a reanalysis of the data found no significant effect for religious affiliation on generosity (Shariff et al., 2016). Thus, in the current review, this retracted study (Decety et al., 2015) due to inaccurate data analysis would be reported with the results of the reanalyzed study (Shariff et al., 2016) (Table 3).

Table 3. Quality assessment.

Demographic variables

Age and gender were considered as important demographic variables in most studies in this review. Three studies did not consider age in their analysis (Kingston & Medlin, 2005; Pandya, 2017; Panofsky, 1976). In one study, age was not included in the model since it was not significantly associated with prosociality (Crosby & Smith, 2015). Only one study did not include the gender variable in their analysis (Robbins et al., 2016).

Participant demographics

Age had a key role in the resulting association. The studies reporting a positive association had a higher mean age (between 8.29 and 10.9) than the studies reporting no association (mean age between 3.45 and 5.04 in one study to 9.87). Moreover, Children aged 5 were more prosocial than those aged 3 (Robbins et al., 2016). In fact, this study did not find an association of religiousness to prosociality because the age of the children was more important than their religiousness. Tibetan three-year-olds in the reviewed study behaved the same as three-year-olds in different cultures with different religions (Rochat et al., 2009). Age was also an important predictor in four other studies. There is an association of religiousness to prosociality for older children (mean age 10–12) when the recipient was needy (Sabato & Kogut, 2018). Age was only a significant predictor of outgroup attitudes with older children having more negative attitudes towards outgroups (O’Driscoll et al., 2021). The children’s ingroup symbol preference and negative outgroup attitudes mediated the effect of age on outgroup prosociality (Shamoa-Nir et al., 2020). The increase in prosociality with age resulting from these studies is consistent with other findings (Köster et al., 2016).

Participant demographics: gender

Gender is another variable that was of interest in the studies reviewed. Girls were more often higher on prosociality (Crosby & Smith, 2015; Kingston & Medlin, 2005; Panofsky, 1976), except for only one study in which boys were higher in prosociality (Pandya, 2017). Girls empathized more with children having similar interests to themselves than those having dissimilar interests (Panofsky, 1976). One study found no gender difference in the interaction between gender and other variables (Sabato & Kogut, 2018) and gender as a control variable was not significant in two other studies (O’Driscoll et al., 2021; Shamoa-Nir et al., 2020). Boys have significantly lower levels of church support which predicted lowers levels of spirituality and lower prosociality (Crosby & Smith, 2015).

Study design

Across the included studies, four design features emerged as a potential source of variability (Table A1, last column). These design features include: (1) measuring the different types of schooling/program and comparing home-schooling to public schooling or a spiritual education program, (2) testing mediators/moderators, (3) controlling for moral and cognitive reasoning, and (4) measuring prosociality towards different targets.

First, some studies strategically varied school types. For example, in one study, home-schooled Protestant children were more prosocial than public school children with different religions (Kingston & Medlin, 2005). Another study varied school instruction around religion (Pandya, 2017); teaching children spirituality improved their prosocial scores compared to children who did not follow the program.

Second, three studies proposed and tested mediation models, trying to unpack the association between religiousness and prosociality. For example, spirituality partially mediated the association of religiousness to prosociality; the authors argue that children who receive love, empathy, caring, trust, a sense of community, and the feeling of family from their church are more likely to internalize the church’s values regarding prosociality (Crosby & Smith, 2015).

Third, related to participant demographics like age, two studies explicitly control for children’s moral and cognitive development. Kingston and Medlin (2005) found that Protestant home-schooled children had a slightly higher moral reasoning and altruism compared to public-schooled children. Robbins and colleagues (2016) found that children who passed the false-belief task were significantly more prosocial than those who failed the task. Despite critiques of the measures used to assess moral and cognitive development, these two studies suggest that controlling for these factors, along with child age, may be important to understand the association between children’s religiousness and prosociality.

Finally, four of the studies varied the target of prosociality (see Moran & Taylor, 2022 for a review). For example, two studies measured prosociality towards religious outgroup members (O’Driscoll et al., 2021; Shamoa-Nir et al., 2020). In a social context of conflict between the Muslims and Jews, religiousness shaped the negative attitudes towards the outgroup (Shamoa-Nir et al., 2020). Besides other religious affiliations, the target to prosociality was the experimenter with the same religious background or an inanimate object (Robbins et al., 2016), and fictional needy and non-needy children with a similar background (Sabato & Kogut, 2018).

The demographics and these four design features reflect the context in which the data were collected and caveats which lead to the relationship between religiousness and prosociality. Important controls to the association between religiousness and prosociality were gender, age, type of schooling, spirituality, moral and cognitive development, and targets of prosocial behavior.

Discussion

First, the current systematic review aimed to identify how religiousness and prosociality are measured. Studies relied heavily on parental religiousness. More robust measures of religiousness included multiple dimensions rather than just religious affiliation. Most of the studies reporting no association used parental reports reporting parental’s religiousness. For prosociality, more robust measurement went beyond only parental or child reporting. Second, we examined the association of religiousness to children’s prosociality, with four of nine studies reporting a positive association, having older children than the studies reporting no association. Third, this overall association varied based on study quality, demographics, such as child age and gender, and more importantly study design. A meta-analysis on religious priming and prosociality in adults found 93 studies (Shariff et al., 2015), while a systematic review on religious attitudes and prosocial behavior in adolescents and adults found 100 studies (Oviedo, 2016); given that this review found only nine studies on this topic in childhood, there is a need for more research during this developmental period, and in early childhood in particular since only four of nine studies focused on age 3–5.

Measures of religiousness

Religion is a complex construct and must be measured as multidimensional phenomenon (Kucukan, 2005). There are over hundred measures of religiousness (Hill & Hood, 1999), with no single standard measure (Hall et al., 2008). For this review, religiousness was classified by the dimensions being measured, ranging from unidimensional approaches (e.g., using only religious affiliation) to combining dimensions such as private religious practice, religious belief, and the consequences of religiousness.

First, unidimensional constructs, such as religious affiliation or religious attendance, did not provide any nuance about the religiousness of participants. For example, participants may attend a religious function, but not actively participate or believe (Ahmed, 2009). In this review, two studies that found no association of religiousness to prosociality used affiliation only. This pattern of findings reenforces the need to go beyond religious affiliation or attendance to achieve a multidimensional measure of religiousness.

Second, private religious practice was measured as a count variable (e.g., how often a child prayed), similar to religious attendance, but always alongside other dimensions, such as religious beliefs. This focus on private religious practice in childhood more as a count variable, rather than the content, is a missed opportunity. As they grow, children start considering prayer as private speech, indicating that the form and content of prayer change with age (Spilka & Ladd, 2013). Future research with children should consider how to assess the qualitative aspects of private religious practice while using it in conjunction with other dimensions.

Third, religious beliefs were measured through the child’s understanding of the supernatural, rituals and dogma (Huber & Huber, 2012). Components of religious beliefs, such as viewing God as an authority figure, predicted prosociality (Bayramoglu et al., 2018). A growing body of literature involving religious priming on adult participants indicates a robust effect on prosociality (Shariff et al., 2015). This systematic review reaffirms these findings; three out of four studies that used religious belief found a positive association with children’s prosociality. Results are also indicative that religious belief is an important dimension that needs to be measured with more than just one item using reliable measures.

Fourthly, three studies measured the consequence of religion. If religiousness requires more than cognitive processes, such as religious beliefs, through experiencing religion and its consequences, religion becomes a way of life (Huber, 2007). The “consequences of religion” offers a more robust measure of religiousness than religious affiliation, attendance, or practice. When religiousness was measured through consequence as part of a model involving outgroup attitudes as a moderator, the results give us a different story (refer to the discussion of the association in O’Driscoll et al., 2021’s and Shamoa-Nir et al., 2020’s study for more detail). In sum, this systematic review suggests that religiousness should be measured through more than one dimension and that more robust measurements (e.g., beliefs, consequences) are more likely to be associated with children’s religiousness.

Finally, religiousness was primarily assessed through parental religiousness and when the children were less than 10 years old, the child report was administered by an adult. Relying solely on parents was another missed opportunity given the range of options to understand children’s religiousness, such as the concept of God (Barrett et al., 2001; Nyhof & Johnson, 2017), the meaning of prayer (Bamford & Lagattuta, 2010; Phelps & Woolley, 2001), and the recognition of religious symbols (Connolly et al., 2002; Taylor et al., 2021).

Prosociality measures

Prosociality was assessed through three constructs: general prosocial behavior, empathy, and altruism. Only one study found a difference between these constructs. Panofsky (1976) found no religious difference for empathy, but a significant difference for generosity. In that study, being Jewish predicted more generosity towards someone similar then someone dissimilar. However, this difference was based on a unidimensional measure of religious affiliation. In sum, there are variations in the robustness of how prosociality was measured—such as multiple reporters, or quasi-experiments.

First, parent and/or child reports were used for the three constructs. Only three studies used only reporting, the others combined it with quasi-experimental measures. As with all self-reports, this type of measure may be limited to assessing only behavioral intentions rather than actual behavior. These findings from this review support the direction that self-report measures of prosociality should be combined with other behavioral measures.

Quasi-experiments may be more indicative of actual behaviors, yet not all experimental approaches yield consistent findings (Franzen & Pointner, 2013). Using different versions of the dictator game yielded mixed results: two found no association (Decety et al., 2015/Shariff et al., 2016; Robbins et al., 2016), and one found a positive association (Sabato & Kogut, 2018). Four studies used more realistic prosocial measures, such as everyday stories (O’Driscoll et al., 2021) or sharing stickers (O’Driscoll et al., 2021; Shamoa-Nir et al., 2020), and candy (Sabato & Kogut, 2018); sharing of peanuts and helping the experimenter (Panofsky, 1976), nevertheless no association was discovered in three of these studies. This set of findings suggests that variations in prosocial measures do not impact the association.

Association of religiousness and children’s prosociality

The findings of this association presented a lot of variabilities. For example, of the four studies that reported a positive association, three used multidimensional religiousness measures, two measured children’s religiousness, while three used child report and one study used a quasi-experimental measure of prosociality. Based on these results, we suggest that researchers prioritize a multidimensional measure of religiousness, measure children’s religiousness, and include more robust measures of prosociality should space permit, with a focus on child-friendly behavioral measures.

Moreover, nuances for this association, such as age and targets of prosociality, are worth noting. For instance, for Sabato and Kogut’s (2018) study this association is conditional on the neediness of the recipient and the age of the participant, such that needy recipients received more candy, and older children gave more candy than the younger participants. Panofsky’s (1976) study found similarity of targets of prosociality effected prosocial behavior in a different ways to different groups. Jews shared more, and Catholics helped more, when the targets of prosociality shared the same interests than when they did not share the same interests. These studies show the importance of focusing on targets of prosocial behaviors and comparing children’s age groups.

In Crosby and Smith’s (2015) study, spirituality mediated the relationship between church support and prosociality. However, given that both constructs of spirituality and church support are measured through the dimension of religious beliefs, this mediation shows the importance of beliefs transmitted through the church’s teachings for children to learn to act prosocial (e.g., Cuadra-Martínez et al., 2019). Despite the positive association between children’s reported religious beliefs (measured through spirituality) and prosocial behaviors, Crosby and Smith’s study (2015) is limited by only self-report prosociality. This reservation holds for two other studies (Kingston & Medlin, 2005; Pandya, 2017), since prosociality was also measured through parental and child reports only, not through actual behaviors.

The association between religiousness and prosociality points to important conditions which need to be included to improve the validity of the study. The association between religiousness and outgroup prosociality, in particular, was found to be mediated by attitudes in two studies. Shamoa-Nir et al. found that the more aware of the ingroup symbol preference (religiousness), the more the negative outgroup attitudes and the less children were willing to give to the outgroup. O’Driscoll et al.’s study found that when religiousness was low (lower essentialist beliefs), children had more outgroup empathy which related to more positive attitudes towards the other religion, leading to more prosocial behavior towards the outgroup. These conditions suggest that when studying this association, research should specify targets of prosocial behavior, measure attitudes towards targets of prosocial behavior, and measure prosociality through actual behaviors, not intent or self-report.

Potential moderators and variability across studies

Study quality

Studies in this review varied in their quality, including small, convenience samples (e.g., Robbins et al., 2016) as well as a retraction following a re-analysis of the data (e.g., Decety et al., 2015/Shariff et al., 2016). More consistent reporting of ethical procedures and the push for open science practices (Open Science of Religion, 2021) are welcomed changes that will improve the quality of future research on this topic.

Participant demographics: age and gender

For the association between religiousness and children’s prosociality, children’s age and gender were important demographic variables explored across the studies. Many of the studies found a positive association between age and prosociality, consistent with previous research. This resulted in studies with older children (average 8–10 years) to find an association between religiousness and prosociality. At the same time, there was only one study in early childhood, ages 3–5, being a period key to children’s religious development (Boyatzis, 2005). Regarding gender, while many studies reported higher prosociality among girls, compared to boys, one study suggested that spiritual educational programs may be more effective in improving boys prosociality, compared to girls (Pandya, 2017). Future research should continue to consider participant demographics, both as main effects and as potential moderators of the association between religiousness and prosociality, particularly in educational settings (Tintori et al., 2021).

Study design

Four design features were identified: type of school, testing mediators, controlling for moral and cognitive development, and target of prosociality. Given the prevalence of religious-based education, school type is a factor to consider in future research designs. The integration of social and developmental theories, such as empathy-attitudes-action (O’Driscoll et al., 2021) and social identity development theory (Shamoa-Nir et al., 2020) also help to tease apart the potential underlying mechanisms relating to children’s religiousness and prosociality. Studies measuring moral reasoning suggest higher moral reasoning and more parental reported prosociality for home-schooled children. As the authors of a study observe, parents are not impartial observers of their children, so research should focus on overt behavior rather than parental reports of prosociality (Kingston & Medlin, 2005). Finally, targets of altruism need to be discussed in the context of parochial altruism. Findings in this review suggest that the social context negative outgroup attitudes and conflict predict lower outgroup prosocial behavior. The importance of changing outgroup attitudes in childhood is echoed by other researchers (Killen et al., 2016). Results of this review suggest that future studies should explicitly note the targets of prosociality to deepen understanding about the link from children’s religiousness and prosociality across contexts.

Limitations and future research

Despite following PRISMA guidelines, grey literature may not have been captured in the search process and one paper was not able to be included in the full-text screening. Unfortunately, most of the studies had a sample of less than 100 participants with mostly WEIRD samples that could have affected the measures, study design and results (Henrich et al., 2010). Yet, the narrative synthesis provides a detailed understanding of the overall association of religiousness to children’s prosociality. The systematic review also notes how the study quality, demographic variables, and design features may influence this association. Complementing these findings, the review also points to three directions for future empirical research. First, given the previous focus on parental religiosity and parental report on children’s religiosity, there is a need for more research into children’s religiousness collected from children’s data, with a focus on early childhood. Second, this systematic review points to the value of complementing self-report of both religiousness and prosociality with more robust forms of measurement. Lastly, future research should specify the target of children’s prosociality to more fully understand how this association may influence cross-religious relations.

Acknowledgements

The authors wish to thank Isabelle Nic Craith for her assistance in reviewing articles’ eligibility for this systematic review.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 https://osf.io/74uqw.

2 One study (Rai & Gupta, 1996) was not found. The first reviewer tried to contact one of the authors, but the communication was not successful. Searching for the article through the Interlibrary loan did not yield any positive results, so this study was excluded.

Appendix

Table A1. Table of characteristics of studies, measures, and results.

Study descriptors					Aim 1		Aim 2	Aim 3
APA citation	Age (M, SD, range)	Sample size (% boys)	Country	Religions	Religiousness measure	Prosociality measure	Key findings on the association between religiousness and prosociality	Study variability
Crosby and Smith (2015)	6–13 years: (M = 9.76, SD = 1.44)	279 (50%)	United States	Evangelical	Religious affiliation. Adaptation of the family religious practices measure (Nelson et al., 2010). A measure of adult religious practices (The Fetzer Institute, 1999). “Church support” measure adapted from Battistich et al. (1995) and Hagborg (1998). Spirituality measure adapted from Fisher (2004), the Francis Scale of Attitudes Toward Christianity (Francis, 1989; Krause, 2009).	Parent report—Adaption of the Prosocial Scale of the Strengths and Difficulties Questionnaire (Goodman, 1997); Child report—Child Altruism Inventory (Ma & Leung, 1991).	There is an association between church support and prosocial behavior but not family religious practices. Spirituality mediated this relationship.	Significantly lower levels of spirituality and prosocial behavior for boys since they perceived lower levels of church support. Behavior intentions measured not actual behavior. Prosocial Target: No specific target.
Decety et al. (2015)/Shariff et al. (2016)	5–12 years: M = 8.29, SD = 2.17	1170 (51.4%)	Canada, China, Jordan, Turkey, South Africa, and United States	Catholicism, Muslim, not religious	Duke Religiousness Questionnaire (DRQ) completed by parents (Koenig & Büssing, 2010), which assesses the frequency of religious attendance, private practice, religious beliefs and spirituality of the household.	Quasi-experiment—Dictator Game; Child Empathy Measure—The Questionnaire of Cognitive and Affective Empathy (QCAE) (Reniers et al., 2011); The justice sensitivity inventory (Schmitt et al., 2005)—a parental report.	More sharing by children in non-religious environments than sharing from children in a religious environment. Paper retracted. After reanalysis, they did not find an association (Shariff et al., 2016).	Age predicted sharing behaviors. Study was retracted due to error in the coding of country. Prosocial Target: No specific target.
Kingston and Medlin (2005)	8–11 years; Home school (Protestants): M = 9.9 Public school (Protestant, Catholics, another religious affiliation or non-at all): M = 9.7 SD not reported.	80 (41.25%)	United States	Protestantism, Catholicism and had either another religious affiliation or none at all.	Religious affiliation. Religious Attendance. Parental questionnaire measuring religious attitudes.	Parents’ ratings of their children’s prosocial behaviors; Children’s empathy was measured with the Adapted Interpersonal Reactivity Index (A-IRI)—(Davis, 1980, 1983), adapted for use with children (Litvack-Miller, 1991). Children’s altruism measure developed by Litvack-Miller (1991).	Positive association: Home-schooled protestant children are more prosocial than public-schooled children with different religions.	Girls rated higher in prosocial behavior. Different schooling programs. No direct behavior observed. Measuring intent not actual behavior. Prosocial Target: No specific target.
O’Driscoll et al. (2021)	5–9 years old (M = 7.09, SD = 1.47)	88 (52.3%)	Northern Ireland	Protestantism, Catholicism	Religious affiliation. Measure of essentialist belief about ethno-religious categories of Protestant and Catholic (Diesendruck & Haber, 2009)	Measure of outgroup-directed empathy adapted from Turner et al. (2013); Measure of outgroup prosocial behavior (O’Driscoll et al., 2018).	There is no association.	Age was only a significant predictor of outgroup attitudes with older children having more negative attitudes towards outgroups. Outgroup attitudes mediated the association. Prosocial Target: children of the other community—Catholic or Protestant.
Pandya (2017)	10–12 years: 34% (1114) 10 years old, 46% (1508) 11 years old and 20% (655) 12 years old. M and SD not reported.	3278 (55%)	Australia, Canada, China, Denmark, Egypt, France, Germany, India, Japan, Saudi Arabia, Singapore, South Africa, Sweden, UK, and US.	Christianity, Muslim, Hinduism, Buddhism, and Judaism.	Religious Affiliation. Records of attendance to SEP and self-practice. The Duke University Religion/Religiosity Index (DUREL) as a measure of religious involvement (Koenig & Büssing, 2010). Parental questionnaire measuring parental attitudes toward religion and values.	Self-report altruism scale (Rushton et al., 1981); Prosocial Personality Battery (Penner et al., 1995).	There is an association between religious belief in the form of teaching children spirituality and prosocial behaviors.	Spiritual educational program. Self-practice was an important variable. No actual behaviors measured. Prosocial Target: No specific target.
Panofsky (1976)	7-year-olds. M and SD not reported.	96 (50%)	US	Jewish and Catholic	Religious affiliation through the school.	Empathy task (Feshbach & Roe, 1968); Altruistic tasks (Rubin & Schneider, 1973).	No association between religious affiliation and empathy/altruism.	No association to empathy. Similarity effects Jews when sharing, and Catholics when helping. Prosocial Target: children with similar/dissimilar interests.
Robbins et al. (2016)	3-year-olds (M = 3.45, SD = 1.59) and 5-year-olds (M = 5.04, SD = 0.93).	29 (51.65%)	India	Buddhism	Religious affiliation.	Simple dictator game adapted for the culture (Rochat et al., 2009).	No association of religious belief to prosocial behavior as religious affiliation did not make a difference to the prosocial behavior.	Age is making the difference in prosocial behaviors. Prosocial Target: Experimenter from the same religious background or inanimate object (toy).
Sabato and Kogut (2018)	5–12 years, M = 8.29, SD = 2.17	478 (45.71%: 55.8% Jewish, 35.62% non-religious)	Israel	Judaism, non-religious	School’s religious affiliation.	Dictator game paradigm.	There is an association of religious beliefs to prosocial behavior for the older age group and for a needy target/recipient.	No gender difference. Children with a mean age of 10–12 were more prosocial than children aged 7–9. Prosocial Target: Classmates with a similar background; needy children and non-needy children.
Shamoa-Nir et al. (2020)	6–13 years: Jewish M = 9.87, SD = 2.48; Arab-Muslims M = 9.35, SD = 2.19	387 (43.15%: Arab-Muslim 44%, Jewish 42%)	Israel	Judaism, Muslim	Religious affiliation. Ingroup symbol preference adapted from (Tomovska et al., 2019).	Outgroup prosocial giving (O’Driscoll et al., 2018).	There is no association between ingroup symbol preference and outgroup prosocial giving. Children with more positive attitudes to the other religion, were more prosocial to the outgroup.	The social context of conflict in religious beliefs might have shaped the negative attitudes towards the outgroup. Prosocial Target: ingroup and outgroup children—Muslims and Jews