Free will in the eyes of Muslims and Christians Authors: Ameni Mehrez, Edouard Machery, Valerio Capraro Published: 2026 ABSTRACT AbstractThe tension between human free will, on the one hand, and divine foreknowledge and sovereignty, on the other, has been addressed differently in different Islamic and Christian traditions. This article examines whether these differences are reflected in how lay Christians and Muslims think about free will, control, blame, and punishment, both when answering abstract questions about God and free will, and when making judgments about concrete situations. We find that despite theological differences among Christian denominations, Christians have similar attitudes in both abstract and concrete contexts, suggesting that lay religious experience is not fully penetrated by theological differences. Muslims differ from Christians: In agreement with the “Qur’an,” they embrace both free will and divine sovereignty when answering abstract questions, but they deemphasize free will when making concrete judgments. These findings reveal the complexity of the interplay between theological reflection and lay moral and religious cognition, and show that the tension between free will and divine foreknowledge and sovereignty is not just a challenge for philosophers and theologians, but is part of lay believers’ religious experience.Keywords: religion, Islam, Christianity, free will, divine foreknowledge This paper is the first to examine whether lay believers think differently about free will and divine foreknowledge and sovereignty across several Muslim and Christian denominations in two Christian- and Muslim-majority countries in light of the historical theological differences regarding the problem of theological fatalism. It casts light on how historical differences in theology influence or fail to influence lay religious cognition. Introduction For more than 15 centuries, Jewish, Muslim, and Christian philosophers have debated the compatibility of human free will and divine foreknowledge (i.e. God's knowledge of future events). At the core of this debate lies an inherent tension: If an omniscient God has complete knowledge of future events, including human actions, how can people be genuinely free to do otherwise? This question, also known as “the problem of theological fatalism” (1–3), has driven centuries of thought, from Augustine (4) in the fourth and fifth century AD and Boethius (5) in the sixth century AD to Al-Ghazali (6) and Averroes (7) in the 12th century AD, all of whom tried to alleviate the tension between free will and divine foreknowledge. Christian and Muslim philosophers, since at least Augustine in the fourth and fifth centuries (8), have similarly examined various ways of alleviating the apparent tension between divine sovereignty (i.e. God's omnipotent authority and control over all creation), particularly with respect to predestination (God's omnipotent decision about who will be saved), and free will (9). Despite extensive research on how lay people understand free will in the social sciences, little is known about whether lay religious individuals experience a conflict between free will and divine foreknowledge and sovereignty. Past research suggests that pious individuals believe that God knows in advance how humans will act (10). Do they also believe that they are genuinely free to pursue their actions? Or do they believe they merely follow predetermined and foreseen paths? And most importantly, do they reconcile the beliefs in free will and divine foreknowledge and sovereignty, somehow managing to alleviate the tension between them? This article reports the first attempt to answer these questions using online and in-person survey data from deeply religious individuals across religions and cultures in a highly powered, preregistered study. We examine how Christians, from three different denominations, and Sunni Muslims think about God's foreknowledge and free will in abstract and concrete situations. Abrahamic religious denominations place varying emphasis on free will and divine foreknowledge and sovereignty, each perhaps addressing thereby in its own way the tension between free will, on the one hand, and divine foreknowledge and sovereignty, on the other. Among Muslims, Sunni Islam often emphasizes God's foreknowledge and omnipotence (11); for instance, in the “Qur’an,” the word “alim,” عليم, which is one of God's 100 names, is mentioned 32 times to refer to God All-knowing, a notion related to divine foreknowledge. Among Sunni Muslims, believing in divine decree (قضاء or “qada”) is a core pillar of faith, which entails that everything that happens in this universe happens by the will of Allah, a notion related to divine sovereignty. This idea of unchanging decree is also manifested in the “Preserved Tablet,” اللوح المحفوط, which is a record of all that will happen, which blends foreknowledge and sovereignty. Christians’ attitudes toward the compatibility of free will, on the one hand, and divine foreknowledge and sovereignty, on the other, vary across denominations. We examine Calvinists’ and Catholics’ attitudes because they tend to be opposite to each other as well as Methodists’ attitudes because they are halfway between those. According to Calvinism, God predetermines and as a result foresees who will be saved and who will be damned. As Calvin put it (“Institutes of the Christian Religion,” 1.16.2), “nothing happens but what [God] has knowingly and willingly decreed.” In contrast, Catholicism tends to deemphasize foreknowledge and sovereignty, while emphasizing free will and the responsibility of individuals (although historically there has been a variety of opinions about free will among Catholic thinkers; e.g. 12): The “Catechism of the Catholic Church” always highlights the compatibility of providence and foreknowledge with free will and responsibility. Similarly, while Methodists embrace providence and foreknowledge, they deny that God predetermines which individuals will be saved, and they highlight free will. On the basis of the theological differences between these four Abrahamic religious denominations, we expected lay believers to alleviate the tension between human free will, on the one hand, and divine foreknowledge and sovereignty, on the other, differently across religious denominations: We expected Calvinists and Muslims to put less emphasis on free will, and Catholics as well as Methodists to emphasize it. Past research about believers’ attitudes toward free will (for review, see 13) and related notions is limited. Most people report incompatibilist intuitions about responsibility and determinism in the United States, Hong Kong, India, and Colombia, while also reporting believing that our world is indeterministic, suggesting they embrace free will (14). Similarly, in all countries surveyed in wave 7 (2017–2022) of the World Values Survey (15) people on average agree with the item used to assess belief in free will, although the strength of this agreement varies across countries (for further discussion, see 16–18); this is, in particular, the case in Tunisia and the United States (respectively, M = 6.2 and SD = 2.6 for Tunisia, and M = 7.7 and SD = 2.0, on a 10-point scale with “1” meaning “not at all” and “10” meaning “a great deal”). Developmental studies have found some cultural differences between China and the United States, particularly about how much choice people have and how their actions are constrained by social factors (19, 20). Similarly, Hannikainen et al. (21) report that people in East Asia think about whether a choice is free when the agent could not have acted otherwise, differently from the rest of the world, although they found little cross-cultural variation concerning the incompatibility between free will and determinism. Finally, although there is some work relating belief in free will and religiosity (22, 23), there is to our knowledge little work examining attitudes toward free will across religions and a fortiori across religious denominations. There is more cross-religious evidence about attitudes toward fate (the idea that future events will necessarily happen independently of one's actions), although unfortunately past research rarely distinguished believers by denominations, aggregating rather at the level of religions. According to a 2011–2012 PEW study, 70% of Muslims said they believe in fate in 19 out of the 23 countries surveyed (24), with percentages ranging from Tunisia (98%), Egypt (93%), and Jordan (91%), to Albania (44%) and Kosovo (50%), all overwhelmingly Sunni Muslim countries. Muslims are also more fatalist than Christians with respect to poor health (25–27). However, in a survey conducted in Egypt, an overwhelmingly Sunni Muslim country, in the 1960s, 90% of peasants believed that people earn their social positions by their own work and efforts (28). Indonesian Christians (two Indonesian Christians out of three are Protestant for one Catholic) also appeared to be more fatalist than Indonesian Muslims, who are overwhelmingly Sunni (29). While Lebanese Christians (including Maronite Catholics and Greek Orthodox) appeared less theologically fatalist than Lebanese Muslims (who could be either Sunnis or Shia), both groups “simultaneously feel equally empowered on a personal, individual level” (29, 1724). In addition to being inconclusive, past research often aggregates various religious traditions within Christianity and Islam, contrasting Christians in general with Muslims in general. This paper disaggregates these traditions, distinguishing Catholics, Methodists, and Calvinists, and focusing on Sunni Muslims among Muslims. To examine whether lay believers vary in their attitudes toward free will across Abrahamic religious denominations (with Sunni Muslims and Calvinists deemphasizing free will, and Catholics and Methodists emphasizing it), we presented participants in the United States (Sunni Muslims, Calvinists, Catholics, and Methodists) and in Tunisia (Sunni Muslims) with two types of materials: scales measuring attitudes toward constructs such as free will and divine sovereignty and vignettes describing concrete situations (see Tables S1 and S2 for descriptive statistics and attrition rates, respectively). Past research suggests that judgments about free will and responsibility vary when people are asked to make a “concrete” judgment about a crime committed by a particular individual (14, 30, 31) and when asked “abstractly” about free will (32), a finding widely replicated (33), including across cultures (14). It is, however, unknown whether it is robust across religious groups. Participants (n = 1,571) were randomly assigned to one of two conditions, which varied whether the agents described in vignettes had to complete an action that is unethical according to commonsensical standards that are widely shared across societies, under low or high duress. Participants read three vignettes presented in random order. Each vignette was followed by six questions in a fixed order. The first question was a trivial yes/no comprehension question; the second asked whether the agent was free, the third whether she was in control, the fourth whether she had to do what she had done; the fifth whether she should be blamed; and the sixth whether she should be punished. Questions 2–6 were all followed by 5-point Likert scales ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). Figure 1 illustrates our research design. Experimental design. The following vignette is drawn from the low duress condition: Sofia studies history at the university. Her acquaintance, who also studies history, asks her to help him during the exam. The acquaintance tells Sofia that he is not going to talk to her anymore if she does not help him during the exam. Sofia decides to help him during the exam. The high duress vignette replaces the third sentence with the following: “The acquaintance tells Sofia that he is going to spread gossip about her taking drugs if she does not help him during the exam.” We expected high duress to matter more for free will, had to, and control, less for blame and punishment. Punishment and blame plausibly depend not only on whether the action is done under duress but also on what is done and on the results of the action, which are independent of duress (34). Participants were then asked to complete three scales in a fixed order (further details about the scales and measurements can be found in the Materials and methods section): the DUREL (35) or a modified version of the DUREL if they were Muslims, the multidimensional fatalism measure (MFM) (36), and the Free Will subscale of the Free Will and Determinism Plus scale (FAD Plus; 37). The DUREL is a 5-item questionnaire measuring three aspects of religiosity (35): “organizational religious activity (DUREL 1), non-organizational religious activity (DUREL 2), and intrinsic religiosity (DUREL 3).” The MFM measures five constructs—the beliefs in Divine Control (e.g. “Everything that happens is part of God's plan”), Luck (e.g. “When I get what I want, it's usually because I’m lucky”), Helplessness (e.g. “I feel that nothing I can do will change things”), Internality (e.g. “I feel that when good things happen, they happen as a result of my own efforts”), and Fatalism (e.g. “I have learned that what is going to happen will happen”)—whereas the Free Will subscale of the FAD Plus (e.g. “People have complete control over the decisions they make”) measures the belief in free will. The subscale Divine Control captures aspects of both attitudes toward divine foreknowledge and divine sovereignty. Participants were thus presented with a total of six subscales. Measurement invariance of the Free Will subscale from the FAD Plus and of the five subscales of the MFM was assessed, showing configural and metric invariance across our religious denominations for each subscale (see the Materials and methods section). Participants finally answered some demographic questions as well as an attention check at the end of the survey. We set the significance level at 0.005, and we treat P-values between 0.05 and 0.005 as suggestive and in need of further confirmation (38). Our analysis plan was preregistered on Open Science Foundation (OSF) (osf.io/4fe62; all deviations reported in the Materials and methods section). Results Abstract scales We first examined Muslims’, Catholics’, Calvinists’, and Methodists’ abstract judgments about free will, divine control, internality, luck, helplessness, and fatalism to test the preregistered hypothesis that Catholics and Methodists have a stronger belief in free will compared to Muslims and Calvinists. We computed regression models predicting each of the subscale variables (Divine Will, Fatalism, Free Will, Luck, Helplessness, and Internality) and using Catholics as a reference category (see Tables S3 and S4 for the full regression estimates). Figure 2 reports the means for each subscale and each denomination. Figure 3 reports the regression coefficients across the six subscales for the four religious groups with Catholics as a reference group. Means for the six subscales per religious denomination. Regression coefficients for the six subscales for the four religious groups (reference group: Catholics). Blue shading indicates negative coefficients, and red shading indicates positive coefficients, with color intensity proportional to coefficient magnitude. Asterisks denote statistical significance levels (*suggestive, **significant). First, contrary to our preregistered hypothesis, there is not much variation among the Christian denominations. Calvinists’ and Methodists’ belief in free will did not significantly differ from Catholics. Similarly, Calvinists did not seem more committed to Divine Control (God's influence over one's life) than Catholics. Calvinists and Methodists are significantly less likely to believe that life events are due to luck than Catholics; this rejection of the role of luck in life is, at least for Calvinists, consistent with their emphasis on predetermination. What is more surprising, Tunisian Muslims expressed stronger belief in free will than Catholics; they also express a significantly stronger belief in Internality than Catholics—viz. the belief that life is determined by one's own actions. In addition, both Tunisian and US Muslims appear, as is to be expected, to be more committed to Divine Control and Fatalism than Catholics. These results are not affected when we control for participants’ scores on the three DUREL factors and age (Tables S3 and S4). In addition to testing the preregistered hypotheses, we also explored the correlations between the five subscales of the MFM, the Free Will subscale of FAD Plus, and the three subscales of the DUREL. Figure 4 summarizes the correlations for the five samples (Tables S5–S10). Correlations between all subscales for the five religious groups. Panel A: Calvinists; Panel B: Catholics; Panel C: Methodists; Panel D: US Muslims; Panel E: Tunisian Muslims. Similarities across the five samples, despite theological differences, are noticeable. For all denominations except Tunisian Muslims (r = 0.1; arguably because of the low variance in intrinsic religiosity among Tunisian Muslims), intrinsic religiosity (DUREL 3) significantly correlates with a greater belief in divine control (0.22 < r < 0.51; see also 39); it also significantly correlates with a greater belief in free will (0.16 < r < 0.21), except, as expected, among Calvinists (0.11) and, in contrast to US Muslims (0.16), Tunisian Muslims (0.04), perhaps again because of low variance. Perhaps surprisingly, the belief in free will correlates with the belief in divine control (0.15 < r < 0.23), except among Tunisian Muslims (0.04). Concrete scenarios We then examined how Muslims, Catholics, Calvinists, and Methodists assign free will in concrete situations (see Fig. 5 for the means and confidence intervals for each scenario). Means and 95% CIs for each DV, condition, and group. Panel A: Free Will; Panel B: Control; Panel C: Had to; Panel D: Blame; Panel E: Punishment. To analyze answers to the vignettes, as preregistered, we fitted three nested regression models, including dependent variable (DV) (five levels, free will as reference category), condition (two levels, high duress as reference category), and their interaction as fixed factors as well as vignette and participants as random factors (varying the intercept); then, we added religious group (five levels, Catholics as reference group) and the interaction between religious group and condition; then, we added the three-way interaction between religious group, condition, and DV. We then compared the fit of these three models by means of ANOVAs. The third model fitted significantly better (Tables S11–S14). Answers to the four DVs differed from the answers to the free will question. Low vs. high duress also influenced the DVs differently: low vs. high duress influenced answers about whether the agent had to act significantly less than answers about free will; in line with our prediction low vs. high duress influenced answers about deserved punishment suggestively less than answers about free will (an effect moderated by a three-way interaction). We did not find any effect for blame, however, which is consistent with recent findings suggesting that blame behaves more like wrongness judgments than like punishment judgments (34). Finally, the amount of duress influenced answers differently across religions, with Tunisian Muslims being significantly less affected than Catholics (the effect is suggestive for US Muslims). In light of the variation across DVs, we examine each DV separately in an exploratory analysis (Fig. 5 and Tables S15–S19), treating condition, religion, and their interaction as fixed factors, and letting the intercepts vary for vignettes and participants. Tables S15–S19 report additional models controlling for various individual-differences measures (intrinsic religiosity [DUREL 3], beliefs in divine control, luck, helplessness, internality, fatalism, and free will). Adding these covariates had little impact on our findings. We review the main findings for each DV in turn. Participants assigned less free will when the agent acted under high duress, but the impact of Duress on free will was significantly smaller for Tunisian Muslims than for Catholics (Table S15). In addition, Tunisian Muslims assigned significantly less free will than Catholics (β = −0.805, P < 0.001; the effect was neither significant nor suggestive for US Muslims, although in the same direction). The finding that Tunisian Muslims assign less free will in concrete contexts contrasts with the finding that Tunisian Muslims report a stronger belief in free will and in divine control than Catholics when reporting abstract beliefs (see the analysis of the abstract scales above). Participants assigned more control when the action was committed under low duress (Table S16); but the impact of Duress was significantly smaller for Tunisian Muslims than for Catholics (a similar interaction was neither suggestive nor significant for US Muslims, although in the same direction). In addition, in agreement with the finding about free will, Tunisian Muslims (β = −1.079, P < 0.001) assigned significantly less control than Catholics (the effect was suggestive for US Muslims, P < 0.05); Methodists assigned more control than Catholics, although the effect was only suggestive. Participants agreed less that the agent had to act when the action was committed under low duress (Table S17), but this effect was qualified by an interaction with religion for Tunisian Muslims. In addition, US Muslims were more likely than Catholics to say that the agent had to act (β = 0.703, P < 0.001), while Tunisian Muslims were less likely (β = −0.742, P < 0.001). Methodists (β = −0.417, P < 0.005) and Calvinists (β = −0.283, P < 0.05) were less likely than Catholics, although less so than Tunisian Muslims, to say that the agent had to act. Participants judged that the agent deserved less blame when she was under high duress (Table S18), although this effect was not present among Tunisian Muslims (β = −0.878, P < 0.001). In addition, in line with the judgments about whether the agent had to act, Methodists and Calvinists significantly judged that the agent deserved more blame than Catholics. Finally, participants also judged that the agent significantly deserved less punishment when she was under high duress (Table S19), although this effect was not present among Tunisian Muslims (β = −0.679, P < 0.001) and was suggestively smaller for Methodists (β = −0.369, P < 0.05) compared with Catholics. In addition, in agreement with the judgment about whether the agent had to act and the blame judgment, Methodists significantly and Calvinists suggestively judged that the agent deserved more punishment than Catholics (Table S19). Discussion Our preregistered study provides, for the first time, evidence about how believers understand free will across religions and cultures. Collecting high-quality data from individuals strongly committed to their faith, we examined whether Muslims in Tunisia and in the United States as well as Catholics, Calvinists, and Methodists in the United States embrace free will similarly in abstract and concrete contexts. Contrary to what we had predicted on the basis of theological differences across denominations concerning the significance of free will and its relation to divine foreknowledge and sovereignty, Christians in the United States reported very similar beliefs about free will in abstract and concrete contexts. Overall, mean answers given to the six subscales used in our study were similar across Christian denominations, as were the correlations between these subscales. In particular, Christians did not appear to differ across denominations when asked abstract questions about free will. We also found few differences in how Christians answer questions about free will and related concepts in concrete situations; in particular, Catholics were not more likely to assert free will and control than Calvinists even when considering concrete situations. While theological views about divine foreknowledge and predestination do differ and while they are extensively discussed in official documents, online and traditional media, sermons, and so on, they might not be central to the religious experience and practices of devoted Christians, at least in the United States. In addition, theological discussions of free will often take place in soteriological contexts in relation to the question of divine grace and may thus not be salient when it comes to judging about the freedom of everyday choices. In addition, contrary to a stereotype about Muslim fatalism (40, 41), Muslims embrace free will in abstract contexts. Tunisian Muslims express a “stronger” commitment to the reality of free will than Christians. At the same time, Muslims report a stronger belief in Divine Control (for US Muslims) and Fatalism (for both United States and Tunisian Muslims) than Christians. American Muslims are also more committed to free will the more religion matters in their life, and the more they assert divine control, the more they embrace free will (two effects not found for Tunisian Muslims, perhaps due to low variance). Thus, contrary to our expectations, Muslims assert both free will and divine control, at least in response to abstract prompts. This finding is consistent with the Prophet's own teachings: The Qur’an asserts both the Divine Decree and human freedom, although it discourages speculations about their compatibility. In contrast to what we observed with abstract prompts, Muslims deemphasize free will in concrete contexts, viz. when making a judgment about a specific action committed by a particular individual: They are more likely to deny that people committing various crimes had free will and control than Christians. Less ascription of free will and control is consistent with our hypothesis that varying emphasis on free will, on the one hand, and divine foreknowledge and sovereignty, on the other, across religious traditions affects lay cognition. We speculate that Muslims behaved differently in abstract and concrete contexts because of the way they bring their beliefs regarding free will, on the one hand, and foreknowledge and divine sovereignty, on the other, to bear on abstract and concrete questions. When they respond to abstract questions, Muslims are not compelled to worry about the compatibility between free will, on the one hand, and foreknowledge and divine sovereignty: As it were, they answer each question separately from the other, agreeing that God decides and knows in advance how humans will behave and subsequently agreeing that humans are free and responsible. In contrast, when they assess a specific action described, in a concrete scenario, as having already happened, the tension between free will, on the one hand, and divine foreknowledge and sovereignty, on the other, cannot be overlooked because free will and divine foreknowledge and sovereignty are about the same action: Participants have to consider the responsibility of the agent for a given action, while believing that God had omnipotent control over this very action and foresaw it. Since the action has already occurred and the character in the vignettes already committed a wrongdoing, Muslims turn toward the default explanation for what has happened: It was preordained and foreseen. As a result, they tend to disregard free will in an attempt to resolve the psychological tension between free will, on the one hand, and divine foreknowledge and sovereignty, on the other. Similarly, the well-known Sourat al Kahf in the Qur’an (18: 60–82) describes how the Prophet Moses challenged the actions of another servant of Allah known as “Al Khidr,” who, Allah told Moses, was even more knowledgeable than him, only to realize that Al Khidr's actions, as arbitrary as they might have seemed, reflected Allah's decisions and wisdom. The lesson is that with our limited knowledge, we humans should not question the predetermined and foreseen events that we encounter, even when they seem wrong or incomprehensible. We note, however, that we did not directly manipulate the saliency of divine foreknowledge and sovereignty in the concrete scenarios and that our explanatory hypothesis remains speculative. Further research should test it directly. The similarity between US Christians and US Muslims when making judgments about blame and punishment and in how duress shapes their judgments is also worth discussing. Both US Muslims and US Christians assigned less blame and punishment under high duress, whereas Tunisian Muslims tended to judge similarly independently of the degree of duress. This pattern suggests that factors beyond religious teachings—such as institutional systems or cultural norms—influence how people assess free will and responsibility. The convergence between US Muslims and US Christians, despite their different faiths, highlights the importance of studying religious groups across diverse cultural contexts rather than within a single setting, as it reveals that religious doctrine alone may not fully explain variation in moral judgment. Our study is limited in various ways. We did not measure participants’ theological knowledge, although we measured their religiosity, and theological differences could influence the free will judgments of believers with a deeper knowledge of theological matters. Our findings about Christians’ and Muslims’ beliefs in free will in abstract and concrete contexts might not generalize to Christians and Muslims in other denominations (e.g. Shia Muslims) or to Catholics, Methodists, Calvinists, and Sunni Muslims in other countries (42). We also do not know whether our results would generalize to Christians and Muslims less religious than participants in our sample. In contrast, we expect our results to generalize to other vignettes contrasting low and high duress situations, provided that participants are tested in a between-subjects manner; our results might not generalize if participants are presented with both low and high duress situations. Finally, the differences between Tunisian and US Muslims should be investigated further to assess whether the differences are due to differences in response styles to surveys (with a greater tendency to use the extreme points of a scale, e.g. 43) or to genuine cultural and religious differences in religious and moral beliefs. To summarize, philosophers and theologians have theorized about the compatibility of free will and divine foreknowledge and sovereignty for centuries, resulting in striking differences among religious denominations. Our study is the first to examine whether these theological differences are reflected in lay cognition by examining how Muslims in two different cultural contexts and Christians from several distinct denominations understand free will. Both when asked abstract questions about free will, fatalism, and divine control, and when making concrete judgments about individuals acting under some degree of duress, Christians tend surprisingly to express similar beliefs despite theological differences across denominations, suggesting that theological differences matter less than similarities of religious practices among American Christians. When they answer abstract questions about free will, Muslims are as committed to free will as Christians (or even more so for Tunisian Muslims) despite their greater commitment to fatalism and divine control (see also 10), a reflection of theological Islamic teachings. In contrast, when they make concrete judgments, Muslims are less likely to assign free will and control to the individuals described in vignettes than Christians, as we had expected based on the idea that intuitively it is difficult to think of actions as free and as foreknown and preordained by God. Materials and methods Participants To estimate the required sample size needed to observe effects within a group, we collected Prolific data from 92 US religious participants, who reported a Christian, Muslim, or Jewish religious affiliation, and we used their answers to estimate the effect of our experimental manipulation on our five measures (Table S24). More than 90% of the sample reported being Christian. Using the R package simR to generate data sets from 200 to 500 by increments of 50, we then estimated that a sample size of 200 participants would be enough to detect the effect of duress, whether the DVs differ from one another, and the interaction between duress and a DV within a single sample with an alpha set at 0.005. To be conservative, we set a target of 300 participants for each religious group, and we were able to reach that target in nearly all cases. The study was approved by the Institutional Review Board at the University of Pittsburgh (STUDY23060151). The study was classified as “exempt,” which means that participants need not be explicitly asked about informed consent. Data were collected from five different religious groups from November 2023 to February 2024: US Calvinists, US Catholics, US Methodists, US Muslims, and Tunisian Muslims. We collected data from both United States and Tunisian Muslims to disentangle religious and cultural variation: If we were to find that Tunisian Muslims differ from US Christians, it could be either because they are Tunisian or because they are Muslims; if we failed to find any difference between US Muslims and US Christians, it might be because of their shared cultural background. Tunisian data were collected face to face in Tunis, Tunisia, by the polling company One-to-One for Research and Polling; participants were interviewed at their home and were not paid. US data were collected online by CloudResearch in exchange for a small amount for participation. All participants reported being above 18, and all US participants reported being born in the United States. In all the samples, we strived to balance gender and, to the extent, it was possible age. US participants were terminated if they did not answer “Methodist,” “Calvinist or Presbyterian or Member of the Congregational Church,” “Catholic,” or “Muslim,” when asked about their religion. They were also terminated if they answered <3 on a 5-point scale in response to the question, “On a scale from 1 to 5 where 1 = not religious at all and 5 = very religious where do you place yourself?” Tunisian participants were not asked whether they are Muslims since this question might be offensive in Tunisia in a face-to-face data collection context. Participants were excluded if they did not complete the survey, if they failed one or more of the four comprehension questions after the vignettes (as preregistered), or if they failed the attention check at the end of the survey (Q239). Participants who failed one of the comprehension questions were not allowed to finish the survey. Table S2 reports the attrition rates by vignette and religious denomination. The attrition rates are small and similar across religious denominations. We do not aggregate the Tunisian and American Muslim data since we collected these two samples to disentangle cultural and religious influences. Table S1 reports the characteristics of the five samples. Stimuli and translations The English survey was translated into Tunisian Arabic and back translated into English by two native speakers of Tunisian Arabic, and the back translation was checked by one of the two authors, who is a native speaker of Tunisian Arabic. American male and female names were used for US participants and Tunisian names for Tunisian participants. We used four vignette pairs, but one pair accidentally differed between Tunisia and the United States, and we removed it from the analysis. The full text of the vignettes can be found on OSF (https://osf.io/h2ng6/overview). Scales The DUREL is a 5-item questionnaire measuring three aspects of religiosity: “organizational religious activity, non-organizational religious activity, and intrinsic religiosity” (35). This scale has been repeatedly validated, including in Muslim contexts (44), and has been used in dozens of studies. We adapted the Arabic translation of the DUREL for Muslim participants in light of the linguistic peculiarities of Tunisian Arabic, and we also modified the measure of nonorganizational religiosity (the second factor of DUREL): we asked how often participants listen or read the Qur’an. The revised version of the DUREL in Arabic and in English can be found on OSF (https://osf.io/h2ng6/overview). The FAD Plus is a 27-item measure of beliefs in free will, scientific determinism, and unpredictability (37). The FAD Plus has been translated into French (45), Polish (46), and Italian (47), Turkish (48), and partially into Chinese (49), although to our knowledge not into Arabic. The French version of FAD Plus removed some items from the English version (including one from the Free Will subscale) to meet satisfying fit indices benchmarks, and the resulting factor structure differed somewhat from the factor structure of the English scale; similarly, the Polish version removed nine items and proposed a three-factor structure (aggregating the two determinism subscales); the Turkish translation appeared to have satisfactory fit with the data with a slightly modified model (48). To our knowledge, the psychometric properties of the scale have not been examined with religious populations, although the measurement invariance has been assessed across Chinese, Japanese, French, and English (50). The FAD Plus measures four different attitudes: belief in free will, the belief in fatalistic determinism, the belief in scientific determinism, and the belief in unpredictability. We selected the 7-item Free Will subscale (e.g. “People have complete control over the decisions they make”) since the second and fourth subscales were already covered by the MFM and since the Scientific Determinism subscale is not relevant for our purpose. Paulhus and Carey (37) and Carey and Paulhus (22) reported a Cronbach's alpha of 0.70 for this subscale (0.69 in 45; 0.68 in 46; 0.75 in 48). The MFM is a 30-item scale with five subscales: Divine Control, Luck, Helplessness, Internality, and Fatalism. The measurement invariance of the scale has been established between English and Spanish (36); furthermore, the measurement invariance of its French translation was also established between a sample of speakers of French from France and Cameroon (39). Ngueutsa et al. (39) also report an average Cronbach's alpha of 0.86 (0.72 < α < 0.96) for the five subscales. Internal reliability We assessed the internal reliability of the five subscales of the MFM and of the free will subscale of the FAD Plus. Table S20 reports the Cronbach's alphas for each subscale and each religious denomination. Reliability was good to excellent for the Divine Control, Luck, and Helplessness subscales, with alphas ranging from 0.787 to 0.876, 0.804 to 0.876, and 0.783 to 0.840, respectively, across all groups (above 0.70, conventionally treated as a satisfactory benchmark; see 51). The few Cronbach's alphas below the 0.7 benchmark are close to this benchmark and agree with Cronbach's alphas reported in the existing literature for the scales we used. The lowest values were observed in the Tunisian subsample, particularly for Free Will (α = 0.602) and Fatalism (α = 0.689), while reliability estimates were generally higher and more consistent across the US denominational samples. Measurement invariance We assessed the measurement invariance of the Free Will subscale from the FAD Plus and of the five subscales of MFM, which has not been examined beyond Spanish, English, and French, using Multi-Group Confirmatory Factor Analysis with Maximum Likelihood Estimation. Fit indices for the configural and metric invariance of the MFM subscales and Free Will subscale are all above the recommended threshold for invariance testing (52); indices for scalar invariance are slightly below the recommended values (Tables S22 and S23). The configural invariance results indicate that the basic factor structure is the same across the five groups. The good metric invariance in the case of the FAD Plus indicates that respondents from the different groups respond to the items in the same way. Supplementary Material Acknowledgments The authors thank Ian Church for constructive feedback during the APA 2025. Contributor Information Ameni Mehrez, Department of Government, William and Mary, Williamsburg, VA 23185, USA. Edouard Machery, Department of History and Philosophy of Science, University of Pittsburgh, Pittsburgh, PA 15260, USA; African Institute for Epistemology and Philosophy of Science, University of Johannesburg, Johannesburg 2006, South Africa. Supplementary material Supplementary material is available at PNAS Nexus online. Competing interest The authors declare no competing interests. Funding This project was made possible through the support of an award from Hillsdale College (grant 1004). The opinions expressed herein are those of the authors and do not necessarily reflect the views of Hillsdale College. Author contributions Ameni Mehrez (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Visualization, Writing—original draft, Writing—review & editing) and Edouard Machery (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Visualization, Writing—original draft, Writing—review & editing) Data availability The complete data set and the code are available on OSF (https://osf.io/h2ng6/overview). Data analysis and deviations from the preregistration Data were analyzed in R using lavaan and lme4. The analysis differed from the preregistered analysis in that the authors used mixed-effects linear regression instead of one-way ANOVA, in line with contemporary best statistical practices. When the authors analyze the DVs separately, they treat the vignette as a random factor, while they had preregistered treating it as a fixed factor. They did not use weights in their pooled analysis since their primary goal was not to estimate population descriptive parameters (53). References Associated Data Supplementary Materials Data Availability Statement The complete data set and the code are available on OSF (https://osf.io/h2ng6/overview).