Psychology by BogdanovaThe Encyclopedia of Psychology for the Artificial Era

RESEARCH DOSSIER

Religious Cognition, Supernatural Agency, and Causal Attribution

Abstract. Religious cognition comprises the representational and inferential processes through which people conceptualize supernatural agents, interpret extraordinary agency, and attribute ordinary or unusual events to sacred causes. This fundamental critical narrative research monograph examines the history of cognitive research on religion, the psychology of intuitive and reflective theology, minimally counterintuitive concepts, agency detection, theory of mind, anthropomorphism, purpose reasoning, supernatural attribution, spiritual presence, and culturally transmitted agent models. It synthesizes original narrative-memory experiments, cross-cultural developmental investigations, perceptual decision tasks, religious-attribution experiments, large comparative studies, preregistered replication evidence, and contemporary experimental findings through 2026 concerning conceptual coherence and the semantics of teleological explanation. The scientific analysis separates representations from belief endorsement, perceptual agency reports from supernatural inference, and event-level causal explanations from moral or existential meanings. Contrary findings on analytic-thinking primes, mentalizing predictors, and threat-triggered detection demonstrate the need for more precise mechanisms and population-valid measurement. The synthesis develops an evidence-calibrated plural-mechanism account of how cognition, culturally supported expectations, social credibility, and explanatory context interact. It identifies experimental and longitudinal tests for unresolved causal relations while preserving a distinction between empirical psychology and metaphysical claims. The study is an original research synthesis based on published scientific literature, without new participant data or an assertion of systematic-review or meta-analytic procedure.

Keywords: religious cognition; supernatural agency; causal attribution; cognitive science of religion; anthropomorphism; intuitive theology; theory of mind; agency detection; minimally counterintuitive concepts; teleological reasoning; religious attribution; analytic thinking; religious belief; cultural transmission; credibility-enhancing displays; spiritual presence; causal explanation

Introduction: Religious Cognition as an Explanatory Research Object

Religious cognition comprises the processes through which persons construct, retrieve, interpret, and deploy representations of gods, spirits, ancestors, invisible powers, sacred intentions, and causally significant events. These representations may be verbally articulated, image-like, narrative, procedural, or contextually inferred. The scientific object of this monograph is the psychological organization of supernatural agency and causal attribution: how a human thinker represents an extraordinary agent, assigns intentions and capacities to it, and relates that representation to an event requiring explanation. Religious affiliation, spirituality, mystical experience, and the longitudinal acquisition of faith are neighboring but wider objects. The distinction matters because a person may employ an agent representation without publicly endorsing the corresponding doctrine, and may retain a religious identity while varying dramatically in the causal explanations used in daily life. Barrett & Keil (1996) Exline & Wilt (2023)

The guiding research questions concern the content and functional form of agent concepts, the conditions under which ambiguous evidence elicits an agent interpretation, the interaction of natural and supernatural explanations, and the contributions of testimony, learning, intuitive inference, analytic reflection, and context. We also examine whether proposed cognitive dispositions generalize across cultures, which findings have survived stronger designs, and how to distinguish experimentally observed performance from a comprehensive explanation of religion. A causal account has to specify more than an association between a reasoning score and a belief report: it must identify the manipulable mechanism, the context in which it operates, and the outcome actually measured. Gervais (2013) Legare et al. (2012) Maij et al. (2017)

Religious agent representations carry descriptive, evaluative, and relational information. A deity described as omniscient may nevertheless be imagined as inspecting one scene at a time; an ancestor may be treated as attentive to family obligations; a malign spirit may be invoked to interpret misfortune; fate or karma can organize an explanation even without a person-like actor. Accordingly, supernatural attribution should not be equated with the detection of a hidden human-shaped agent. Intentional agents, nonintentional sacred forces, cosmological order, moral accounting, and conventional causal mechanisms form a heterogeneous explanatory repertoire. This difference in representational content is central to construct validity and to identifying which proposed cognitive mechanisms an experiment actually tests. Barrett & Keil (1996) Exline & Wilt (2023)

The evidence assembled here is a critical, source-identified narrative research synthesis. It integrates original narrative-comprehension and recall experiments, agency-detection tasks, cross-cultural and developmental investigations, religious-attribution studies, surveys, registered replications, and theoretical and methodological analyses. The work introduces no new participant data and claims neither systematic-review coverage nor a pooled meta-analytic estimate. Evidence is weighed by design, construct mapping, cultural sampling, preregistration where available, and the extent to which contrary findings have been tested. Philosophical questions concerning the truth of religious claims remain distinct from the empirical question of how humans represent and explain them. Boyer (2003) Sanchez et al. (2017) Van Leeuwen & van Elk (2019)

Conceptual Foundations: Religion, Agency, Causality, and Representation

Agency attribution refers to treating a source of behavior or events as an actor with dispositions, goals, knowledge, or intentions. Mind perception refers more broadly to assigning mental capacities such as experience or agency; theory of mind or mentalizing concerns reasoning about beliefs, desires, knowledge, and perspective. Anthropomorphism is the attribution of humanlike properties to nonhuman or extraordinary targets. Teleological reasoning explains something with reference to its purpose or function, and causal attribution identifies what generated an event. These constructs overlap, but each licenses different behavioral predictions. The scientific need for separate definitions is apparent when a person recognizes an intentional agent yet rejects a supernatural interpretation, or explains an event through karma without imagining a mind that chose the outcome. Epley et al. (2007) Gervais (2013) Kelemen (2004) Exline & Wilt (2023)

A supernatural agent is defined operationally in this monograph as a represented actor to whom participants attribute capacities or modes of existence outside their ordinary expectations about persons, animals, or physical objects. The relevant boundary is local to the conceptual expectations of the population under study. A culturally familiar spirit may have ordinary motives and social obligations while crossing ontological expectations about visibility, death, physical barriers, or spatial location. The concept can be measured through prompted and spontaneous description, narrative inference, memory distortion, judgment under time pressure, and attribution following ambiguous events. Researchers must examine what an instrument presupposes about the agent before interpreting its score as religiosity or a general human disposition. Boyer (2003) Boyer & Ramble (2001) Barrett & Keil (1996)

Supernatural causation is an explanatory judgment that assigns some part of an event's origin, timing, meaning, or trajectory to a supernatural source. It can be proximal, as when a spirit is said to move an object, or distal and indirect, as when divine guidance is inferred from ordinary circumstances. It can also be nested within a biological or social account, for example when illness is treated as a medical process whose occurrence carries sacred significance. Explanatory coexistence means that a respondent accepts multiple causal descriptions, potentially at different levels or for different questions. Evidence of coexistence should never automatically be scored as logical inconsistency: researchers must establish whether competing propositions concern the same mechanism, explanatory level, and time frame. Legare & Gelman (2008) Legare et al. (2012) Exline & Wilt (2023)

Counterintuitiveness must be specified relative to domain-level expectations. In Boyer's framework, a concept becomes cognitively distinctive when it retains most expectations of a familiar category while violating a limited number of them, as with a person who knows hidden thoughts or an artifact that remembers. Such concepts may remain inferentially rich precisely because their unviolated background properties allow the hearer to predict behavior. Minimally counterintuitive concepts are thus a hypothesis about communication and memory, not a complete theory of religious acceptance or a claim that all traditions favor the same imagery. Recall advantage, attribution, ritual commitment, and existential importance require distinct outcome measures. Boyer (2003) Boyer & Ramble (2001) Willard et al. (2023)

Reflective theology denotes explicit verbal or doctrinal descriptions, often shaped by formal instruction; intuitive theology refers to representations deployed in rapid comprehension, imaginative simulation, and everyday reasoning. The labels describe task-dependent forms of representation rather than separate populations of believers. A person can endorse an abstract claim about divine transcendence and still imagine intentional behavior using ordinary person-based constraints. Whether the deviation arises from memory reconstruction, restricted processing resources, the pragmatics of a narrative, or a stable mental model must be tested rather than inferred from the label intuitive. Barrett & Keil (1996) Gervais (2013)

Historical Development of Cognitive Research on Supernatural Agents

Psychological investigation of religious representations developed through several overlapping traditions: anthropological comparison of classifications and spirits, psychology of religious experience, developmental studies of explanation and purpose, cognitive theories of concepts and social inference, and experiments on judgment and memory. The field's contemporary cognitive science of religion consolidated these strands around testable mechanisms. Guthrie's account of anthropomorphism emphasized the ecological importance of interpreting ambiguous environments as potentially agentive, a thesis that helped motivate later agency-detection research. Such work should be read as a historical proposal requiring independent tests of detection thresholds, false alarms, and generalization, rather than as a universal established origin of religious thought. Guthrie (1993) Van Leeuwen & van Elk (2019)

Barrett and Keil's 1996 experiments operationalized an influential distinction between participants' professed theological concepts and their performance in processing stories about God. Three story-processing tasks elicited humanlike inferences and narrative transformations inconsistent with some stated theological commitments. Increasing attention to theological beliefs changed the degree of anthropomorphization, showing that representation depends in part on task framing and cognitive access. The result supplied a particularly useful program: compare explicit statements with memory and inference under specified conditions, rather than treating a single questionnaire as the whole religious concept. Its university-student sampling nevertheless limits claims about all traditions and social settings. Barrett & Keil (1996)

Boyer and Ramble's 2001 work tested whether certain violations of intuitive conceptual expectations influence narrative recall across different populations. Their five free-recall experiments used constructed materials in France, Gabon, and Nepal rather than relying exclusively on familiar religious stories. This design helped separate a hypothesized memory constraint from familiarity with a particular religious tradition. The significant methodological contribution is the use of standardized stimulus structures and multiple cultural settings; the theoretical question of how memorable concepts become socially authoritative remains an additional matter requiring transmission, credibility, and institutional evidence. Boyer & Ramble (2001) Boyer (2003)

Early twenty-first-century research expanded toward individual differences in mind perception, anthropomorphism, purpose explanations, and reflective reasoning. Three-factor anthropomorphism theory related judgments to accessible human knowledge, motivations for effectance or prediction, and needs for social connection. Other programs examined whether mentalizing predicts belief in personal gods, whether children spontaneously explain nature as designed for a purpose, and whether analytical cues transiently shift self-reported belief. These programs produced increasingly heterogeneous evidence and motivated larger cross-cultural datasets and preregistered replications. Their historical progression demonstrates why a psychological explanation of religion requires interacting cognitive and social variables rather than a search for one universal defect or module. Epley et al. (2007) Norenzayan et al. (2012) Kelemen (2004) Gervais & Norenzayan (2012) Sanchez et al. (2017)

Recent work has redirected attention to learned expectations, experience cultivation, culturally supported models of mind, and the evidentiary limitations of the hyperactive agency detection hypothesis. The Interactive Religious Experience Model organizes interactions among prior beliefs, attention, context, bodily states, and experience; work on supernatural attribution emphasizes that a sacred explanation often concerns the meaning of ordinary events rather than a visibly supernatural intervention. Cross-cultural and preregistered studies have subjected simple detection and mentalizing accounts to stronger tests. The intellectual history is therefore one of specification: progressively identifying precisely what changes in a representation, which population is involved, and what kind of causal inference the task justifies. Van Leeuwen & van Elk (2019) Exline & Wilt (2023) Szymanek et al. (2025) Maij et al. (2017)

Methodological Architecture and Standards of Evidence

A study of religious cognition needs a clearly defined target construct, an observable task, and a mapping between the two. Narrative recall measures which propositions are reconstructed after processing a story; it does not by itself measure endorsement of the story's truth. An ambiguous-motion detection task measures classification of perceptual input under particular instructions; it cannot automatically establish the cause of supernatural belief. A supernatural-attribution questionnaire samples explicit explanation of described or personally experienced events; responses depend on wording, presumed audience, and available categories. Each measure may illuminate one component of a broader system, but mechanisms should be inferred only after manipulating relevant causes or triangulating independent measures. Barrett & Keil (1996) van Elk (2013) Exline & Wilt (2023)

Randomized experiments on cueing, cognitive load, stimulus ambiguity, perceived threat, or social framing permit causal inference about the manipulated proximal factor under the tested setting. They ordinarily do not establish the long-term origin or truth of a religious commitment. Surveys establish covariation but face reverse causality, shared-method variance, and confounding by religious upbringing, education, local norms, and cultural expectations. Longitudinal observations improve temporal interpretation when they repeatedly measure cognitive mechanisms and belief practices, but directional claims also require attention to selection, attrition, and time-varying confounding. Cross-national comparisons demand comparable translations and item meaning before estimating differences between societies. Maij et al. (2017) Stagnaro & Pennycook (2025) Richert & Corriveau (2022)

Signal detection designs can separate sensitivity to actual agents from a response bias toward reporting agency, provided ground-truth stimuli and false-alarm opportunities are defined. Causal-attribution experiments should vary a candidate mechanism while holding event valence, likelihood, severity, controllability, and culturally expected interpretation as constant as possible. In conceptual memory experiments, researchers need matched narrative length, emotional intensity, familiarity, and inferential richness across intuitive and counterintuitive concepts. In theology experiments, explicit doctrinal familiarity and language proficiency should be recorded so that an anthropomorphic interpretation is not mistaken for simple unfamiliarity with specialized religious vocabulary. Boyer & Ramble (2001) Barrett & Keil (1996) Ramsay et al. (2023)

Replication is particularly important because effects in this field are often context-sensitive and initial samples may be small. The 2017 preregistered multisite replication of a prominent analytic-thinking prime reported little to no expected effect; its estimate was d = 0.07 with a 95% confidence interval from −0.12 to 0.25 across 941 participants. This is strong evidence about that experimental priming procedure, while leaving open other relations between reflective reasoning, religious socialization, and belief. A sound synthesis identifies the manipulated variable and the study's actual population before speaking of a refuted or corroborated general mechanism. Gervais & Norenzayan (2012) Sanchez et al. (2017)

The present evidence appraisal distinguishes five inferential grades: conceptual plausibility; observational association; experimentally demonstrated proximal effect; convergent independently replicated evidence; and explanatory models that integrate findings across levels of analysis. These grades are analytic descriptors rather than numeric quality scores or formal risk-of-bias ratings. Where only a published abstract has been consulted, the synthesis stays within the outcomes and procedures reported there. Source selection is purposive across the major research programs and their counterevidence. No systematic database search flow, registered review protocol, exhaustive exclusion ledger, or meta-analysis has been performed. Willard et al. (2023) van Elk (2022) Sanchez et al. (2017)

Supernatural Agent Concepts and Intuitive Theological Reasoning

Humans can reason about entities whose reported attributes exceed familiar biological and physical limits because extraordinary properties can be combined with ordinary category knowledge. A deity may be omnipresent yet purposeful, a ghost may lack a material body yet remember relationships, and a spirit may be invisible yet able to communicate. When the majority of ordinary assumptions about agency remain available, the representation supports prediction and narrative inference. The ability to represent an extraordinary agent therefore depends on general conceptual combination and social inference, with religious conventions specifying which unusual properties matter. Cognitive ease in one task may coexist with conceptual strain in another, particularly when an abstract doctrine is translated into a spatially sequential story. Boyer (2003) Boyer & Ramble (2001) Barrett & Keil (1996)

Barrett and Keil's experiments are especially instructive because a narrative can expose default inference patterns that are hidden by explicit verbal assent. Participants who affirm a nonhuman, unlimited divine perspective can nevertheless recall a story as though a deity had ordinary limitations on attention, location, or information access. The experiment demonstrates a discrepancy in task performance, and its attentional manipulation shows that professed theology can influence which representation is deployed. It does not establish that the participants secretly rejected their theological doctrine. A cognitive model must explain context-dependent accessibility, working-memory demand, pragmatic expectations in storytelling, and the difference between a task-specific simplification and an enduring personal concept. Barrett & Keil (1996)

Intuitive representations of agents include goals, intentions, communication, and knowledge. Formal doctrines can modify these expectations, often through explicit instruction and repeated interpretation. A person may acquire a highly abstract concept of divine transcendence while continuing to exploit familiar interpersonal simulations during prayer, story comprehension, or judgment about a perceived intervention. Individual variation in religious knowledge and how often a person encounters particular narrative genres can change which constraints are activated. The empirical prediction is not that every religious concept becomes anthropomorphic, but that instructions, representational format, and demand on attention may shift reliance on person-based inferences. Designs should test this prediction within traditions as well as between them. Barrett & Keil (1996) Epley et al. (2007)

Accounts of minds beyond the body also raise distinctions between representing an agent's beliefs and attributing sensory experience. An ancestor capable of protecting descendants may be represented as possessing intentions without being expected to see in the ordinary visual sense. A moralizing deity may be represented as knowing violations while being described as free of human emotional limitations. Instruments that combine these capacities into a single anthropomorphism score can conceal theoretically relevant dissociations. Mind-perception research invites researchers to ask separately who can act, who can know, who can feel, and which culturally taught expectations govern these capacities. Such distinctions improve comparisons between belief in personal gods, impersonal sacred forces, and nonreligious uses of agency language. Gervais (2013) Norenzayan et al. (2012) Exline & Wilt (2023)

Minimal Counterintuitiveness, Memory, and Cultural Transmission

The minimally counterintuitive hypothesis proposes that a limited violation of intuitive category expectations can make a concept distinctive while preserving enough ordinary structure to support understanding and recall. A person capable of passing through walls is easier to integrate into a conventional narrative than an entity possessing an unconstrained collection of unrelated violations. Boyer and Ramble tested domain-level conceptual breaches with constructed narratives, asking whether recall differs from control materials that lack the targeted violations. Their cross-cultural use of France, Gabon, and Nepal supplied early evidence that conceptual structure could matter beyond one familiar population. The experimental outcome concerns remembered concepts, not the full distribution of religious beliefs in each society. Boyer & Ramble (2001)

Counterintuitiveness is sensitive to conceptual coding. A property regarded as extraordinary by one researcher may be an established conventional expectation in a particular religious community; likewise, novelty can increase memory without a specifically ontological violation. Researchers need independent local judgments of familiarity, typicality, and surprise, and should distinguish breaches of basic-level expectations from more general domain expectations. A narrative featuring an impossible animal may also be more visually striking, humorous, or emotionally arousing than its comparison. Matching these dimensions and measuring both immediate and delayed recall makes a memory interpretation stronger. Boyer & Ramble (2001) Willard et al. (2023)

Even when an unusual representation is preferentially remembered, social transmission requires more than individual memory. People decide whether to retell an account, interpret an extraordinary claim as credible, adopt a prescribed practice, and regard a speaker as trustworthy. Credibility-enhancing displays are observed behaviors that make professed commitments appear costly or consequential to the observer; they are candidates for explaining differences in transmission beyond individual reasoning skill. The important distinction is that an intuitively compelling representation can remain an entertaining fiction, while an ordinary representation can become religiously authoritative through institutions, relationships, and repeated social practices. Lanman & Buhrmester (2017) Maij et al. (2017)

A complete account should distinguish recall advantage, learning rate, endorsement, salience in ritual contexts, and long-term cultural retention. These outcomes can diverge. A remembered religious story may not be believed; a deeply trusted doctrine may be recalled only approximately; ritual conventions may stabilize particular language even when the associated causal representation varies by setting. Longitudinal transmission studies can examine who introduces concepts, how they are corrected, which aspects resist transformation, and how instruction changes the gap between explicit and implicit interpretations. The present dossier uses cultural transmission as a mechanism that constrains religious representations, while the neighboring research dossier on religious belief acquisition addresses individual trajectories of conversion, doubt, and revision more extensively. Boyer (2003) Richert & Corriveau (2022) Lanman & Buhrmester (2017)

Anthropomorphism, Mind Perception, and Mentalizing

Anthropomorphism can arise because human social knowledge supplies a convenient predictive model for otherwise unfamiliar actors. Epley, Waytz, and Cacioppo proposed three classes of determinants: the accessibility and applicability of human knowledge, the motivation to understand or control an environment, and the motivation to connect socially. In an explanation of religious agency, these mechanisms offer testable hypotheses concerning uncertainty, loneliness, interpretive familiarity, and the framing of a supernatural figure. Their theory applies broadly to human judgments about animals, nature, machines, and gods, which means a religious-specific conclusion requires evidence that the same predictors operate with religious stimuli and in relevant populations. Epley et al. (2007)

Mentalizing is a prerequisite for certain judgments about a personal agent's knowledge and intention in a conceptual sense, but such dependence does not imply a simple linear relationship between general mentalizing test scores and supernatural belief. Tests of false-belief reasoning, empathic inference, autistic traits, and mind perception capture different constructs, and self-reported religiosity mixes belief, practice, group membership, and cultural expectation. Norenzayan, Gervais, and Trzesniewski reported associations that motivated a hypothesis connecting mentalizing differences with belief in personal gods. The interpretation of such associations must preserve the distinction between ability to represent a mind and willingness to endorse a specific supernatural mind as real. Norenzayan et al. (2012) Gervais (2013)

Maij and colleagues brought stronger cross-national and socially contextual evidence to this question. Across five studies and a large combined sample spanning distinct national contexts, mentalizing measures did not consistently distinguish religious believers from nonbelievers, whereas exposure to credibility-enhancing displays more reliably predicted belief. This does not imply that mental-state reasoning plays no role in composing an agent concept. It indicates that the presence of social-cognitive capacity cannot by itself explain who comes to affirm such concepts, and it identifies cultural learning as a necessary element in broader explanatory models. The methodological consequence is to avoid inferring belief from a general theory-of-mind score or comparing cultural groups without measuring differences in testimony and observed practice. Maij et al. (2017) Lanman & Buhrmester (2017)

Some religious explanatory systems assign causes to fate, karma, sacred energy, or cosmic justice without attributing the event to a deliberating person. In other systems, personal gods may be represented as exceptionally knowledgeable but still relationally responsive. Researchers should therefore collect information about the culturally specific object before applying a standard measure of anthropomorphism. Open-ended causal explanations can complement structured scales by allowing respondents to articulate agency, moral regularity, and metaphysical causation in their own categories. An investigator who forces every explanation into an agent-versus-natural binary risks conflating supernatural causality with personal mind attribution and overlooking impersonal sacred interpretations. Exline & Wilt (2023) Van Leeuwen & van Elk (2019)

Agency Detection, Ambiguity, and Perceptual Decision

Agency-detection theories start from an ecologically plausible problem: an organism interpreting uncertain sensory evidence sometimes benefits from responding to a possible actor before certainty is achieved. Guthrie advanced a broad anthropomorphism account of religion, and later hyperactive or hypersensitive agency detection device proposals argued that false-positive detection might contribute to supernatural thought. These accounts must be formulated as competing predictions about perceptual bias, sensitivity, environmental stakes, prior belief, and interpretation of an ambiguous signal. The fact that a person can see agency in noise establishes a general cognitive possibility; it does not by itself establish a reliable evolutionary pathway from perceptual error to belief in gods. Guthrie (1993) Van Leeuwen & van Elk (2019) van Elk (2022)

Experiments with ambiguous biological motion and related stimuli provide operational opportunities to measure reports of detected agency. Van Elk reported that paranormal believers were more likely than skeptics to perceive illusory agency in an ambiguous perceptual task. This finding is evidence of an association within that operational context, and its relation to conventional institutional religiosity was less straightforward. One should distinguish paranormal belief scales, theistic endorsement, supernatural event attribution, and religious practice rather than assume they index a uniform cognitive trait. Variations in instructions, stimulus composition, and expectations can also shift response thresholds independently of actual visual sensitivity. van Elk (2013)

A signal-detection framework asks how often participants correctly detect agents, miss agents, report agents when absent, and correctly reject nonagent stimuli. Without sufficient genuine targets and foils, increased reports of agency cannot readily be separated from a willingness to use the response category. Experimental threat manipulations should change precisely specified features of a situation while controlling arousal and ambiguity, and a preregistered analysis should state whether the prediction concerns sensitivity or criterion. These distinctions are especially relevant to claims that danger automatically increases supernatural agency detection. A model predicting a universally heightened detection tendency faces increasingly difficult evidence when contextually matched experiments repeatedly report negligible or contingent effects. Szymanek et al. (2025) Van Leeuwen & van Elk (2019)

A preregistered virtual-reality study investigated whether threat would alter agency detection in a controlled environment and found little support for the anticipated threat-related shift. The result underscores the importance of ecological design as well as replication: simulated surroundings create more immersive uncertainty than simple laboratory displays but still do not reproduce every cultural or personal circumstance in which a sacred agent might be inferred. The observed null pattern bears on the manipulated threat condition and the chosen detection task, not the entire psychology of extraordinary religious experience. Interpretation must remain specific enough to preserve a falsifiable causal hypothesis. Szymanek et al. (2025)

An alternative account treats agent attribution as an interaction among prior expectations, attention, perceived patterns, explanatory goals, and socially supplied interpretations. On this view, a believer's expectation can change what ambiguous information becomes salient, while a narrative or communal interpretation can supply a religious meaning after detection. This perspective can incorporate perceptual evidence without making a single sensory module the source of all religion. The Interactive Religious Experience Model frames experiences as dependent on such interplay among context and cognitive processes, inviting longitudinal and within-person research. Clear tests would separately manipulate contextual religious cues, target ambiguity, and explanatory prompts and then examine detection judgments alongside later supernatural interpretation. Van Leeuwen & van Elk (2019) van Elk (2022)

Teleology, Purpose, and Design in Human Explanation

Teleological explanation represents an object or event in terms of an end, function, or purpose. It can be appropriate when analyzing engineered artifacts or intentional actions, while an evolutionary or physical account may better explain the origin of a biological structure or natural event. Developmental psychologists examine when individuals extend purpose-based reasoning beyond clearly intentional design and what constrains that extension. Kelemen's work on children's intuitive theism argued that purpose expectations can make design-oriented explanations of nature cognitively available early in development. This research program poses empirical questions about explanatory preference, instructional context, and developmental change; it does not establish that all children hold a mature doctrine of a creator. Kelemen (2004) Richert & Corriveau (2022)

Purpose judgments about one's life introduce another kind of explanatory target. A person who asks why a loss occurred may seek a meaning, a moral lesson, or a pattern of personal significance, even while accepting a complete account of the physical and social causes involved. Banerjee and Bloom examined the tendency to give teleological interpretations of life events among religious believers and nonbelievers and its links to mental-state reasoning. Such experiments illustrate how personal relevance can shift explanatory goals. Attributions of purpose in autobiographical events should be measured separately from claims that a hidden agent directly produced the event or from judgments that physical phenomena are biologically designed. Banerjee & Bloom (2014)

Experiments on design and purpose must disentangle three questions: whether the respondent thinks an event has a function, whether the respondent believes it was intentionally arranged, and whether the arranger is identified as a supernatural agent. Affirmative responses to one question do not logically or psychologically guarantee the others. A species may have evolved a useful trait without an intention behind its emergence; an accident may acquire personal meaning without being regarded as orchestrated; a religious person may endorse both evolutionary causation and providence at different explanatory levels. Responses should therefore distinguish objective purpose, subjective meaning, inferred design, and agent-specific intervention. Kelemen (2004) Banerjee & Bloom (2014) Legare et al. (2012)

Modern analyses of religious cognition benefit from examining purpose explanations as one resource in a plural explanatory system. The same individual may favor mechanistic explanations in a science classroom and moral or existential explanations when narrating a personal crisis. Such contextual flexibility generates testable predictions about prompts, audience, prior instruction, and emotional involvement. A study that instructs participants to select only one explanation may impose artificial exclusivity. More informative designs ask respondents to rank or combine explanations, identify where they believe one applies, and describe which observations would change their view. Legare & Gelman (2008) Exline & Wilt (2023)

Natural and Supernatural Explanations in Coexistence

Legare and Gelman's research on explanations of illness among Sesotho-speaking communities in South Africa brought empirical specificity to the coexistence of natural and supernatural causal frameworks. Participants sometimes accepted a biological explanation and an account involving bewitchment for aspects of the same event. Their developmental studies are especially valuable because they show that learning a mechanistic explanation need not automatically displace cultural frameworks concerned with interpersonal action, responsibility, or the timing of misfortune. The findings do not imply that every explanation is scientifically equivalent; they demonstrate that actual psychological reasoning can operate at multiple causal and interpretive levels. Legare & Gelman (2008)

Legare and colleagues subsequently reviewed cross-cultural and developmental evidence for explanatory coexistence, challenging accounts that equate educational development with a straightforward replacement of supernatural by natural thought. The important analytic question is how explanatory models are organized. A participant might locate a supernatural cause at the level of who initiated an event, a biological cause at the level of how the event progressed, and a moral account at the level of what the event means. Alternatively, explanations may genuinely conflict about the same mechanism. Researchers must code these possibilities separately and avoid treating every combined explanation as an undifferentiated dual belief. Legare et al. (2012)

Causal attribution in ordinary life often concerns events that are probabilistic and emotionally consequential. A recovery following medical treatment may be attributed to effective care, prayer, providence, or multiple factors. An unexpected coincidence may be described as chance, fate, guidance, or an indication of a personal obligation. Experiments should vary the base rate of the event, the availability of ordinary causal mechanisms, the presence of an intentionally acting human, and the salience of a sacred narrative. Without these controls, differences in supernatural attribution can simply reflect differences in how participants understand the described scenario. Exline & Wilt (2023) Ramsay et al. (2023)

The co-presence of explanations raises a formal causal-inference issue: when an investigator asks what caused an event, the response may report an efficient cause, a counterfactual condition, an explanatory background, or a narrative reason. An attribution that a deity intended a meeting does not necessarily specify which observable process produced it. Likewise, an attribution to fate may express a stance toward unpredictability rather than a claim about a detectable physical mechanism. A good instrument therefore distinguishes perceived causal intervention from meaningful interpretation, and reports their relationship instead of collapsing both into a generic supernatural attribution score. This conceptual differentiation is supported by research on the diversity of supernatural causes people invoke. Exline & Wilt (2023) Legare & Gelman (2008)

The coexistence perspective also supplies a research agenda for culturally sensitive studies. Investigators can elicit participants' own explanatory vocabulary, test how categories are translated, and track how explanations are jointly deployed across settings. Within-person comparison may reveal that people distinguish a doctor's account of infection from a religious interpretation of suffering without experiencing those descriptions as rivals. Other participants may explicitly state that a sacred cause overrides a natural one. The scientific task is to record and model that variation. A rigorous synthesis avoids a single binary variable that would erase meaningful differences in how causal claims are made. Legare et al. (2012) Richert & Corriveau (2022)

Supernatural Attribution, Event Features, and Explanatory Goals

Supernatural attribution research asks which circumstances prompt an individual to see divine, demonic, ancestral, spiritual, karmic, or fateful causation in a particular event. Exline and Wilt organize this field around antecedents and consequences of attributing real events to extraordinary sources. Their synthesis emphasizes that such attributions often refer to an ordinary occurrence interpreted as a manifestation of a supernatural cause, rather than to an alleged suspension of natural law. This difference broadens the proper measurement domain to include daily coincidences, fortunate outcomes, moral choices, suffering, and apparently meaningful patterns, while requiring researchers to respect the different agent concepts named by participants. Exline & Wilt (2023)

Event valence, importance, personal relevance, controllability, surprise, and perceived probability can all influence which explanation becomes salient. A serious life event may recruit a search for meaning; an outcome regarded as morally consequential may invite a judgment about reward or punishment; a highly improbable coincidence may draw attention to possible intentional arrangement. Ramsay and colleagues investigated event properties alongside experimental religious priming in studies of supernatural explanation. Their work illustrates why stimulus design matters: an apparent religious-cue effect can depend on whether the described event already affords a sacred interpretation, and findings about one prime or scenario type should not be universalized. Ramsay et al. (2023)

An attribution may also be socially strategic or relationally expressive. Saying that a good outcome was a blessing can communicate gratitude, identity, or solidarity, while blaming an event on malign spiritual influence can communicate moral concern or an attempt to assign responsibility. Such uses warrant direct study, because questionnaire endorsement may register speech conventions as well as literal causal commitment. Investigators can compare private and public responses, use context-sensitive qualitative interviews, and ask counterfactual questions about what participants would expect if the named supernatural cause were absent. These strategies are especially important where expressing or withholding a religious attribution bears social costs. Exline & Wilt (2023) Maij et al. (2017)

Supernatural explanations can organize coping responses, perceived control, and moral action. The fact that an explanation has a psychological function does not establish whether it improves or harms well-being. For some individuals it may support meaning, gratitude, and tolerance of uncertainty; for others it may increase guilt, fear, or conflict. A causal analysis requires attention to the valence and content of the attribution, the individual's resources, the presence of distress, and the broader social context. The present dossier therefore studies how causal explanations arise and function, while research on religious coping, spiritual struggle, and clinical outcomes addresses their longitudinal and therapeutic consequences in dedicated objects. Exline & Wilt (2023)

Different spiritual traditions may encourage highly personalized divine intention, ancestor-mediated influence, impersonal cosmic causality, or more selective forms of sacred interpretation. Scientific instruments should permit these distinct explanatory units, and should not equate absence of belief in a personal creator with absence of supernatural causal reasoning. Cross-cultural validity requires establishing the relevant local categories before coding answers. Combining qualitative elicitation with experimentally standardized event descriptions can make differences intelligible and avoid the circularity of defining supernatural attribution only by resemblance to the researcher's own preferred religious tradition. Richert & Corriveau (2022) Exline & Wilt (2023)

Spiritual Presence, Absorption, and Predictive Interpretation

Experiences of sensed presence, communication, or divine closeness provide another entry point to the relation between representation and attribution. An individual may report a feeling that someone is present and subsequently identify that presence as a god, spirit, ancestor, or secularly interpreted companion. Psychological research should distinguish the occurrence of the experience, its sensory qualities, the immediate interpretation, and the longer-term commitment attached to it. Each can vary independently. An account centered exclusively on detection errors overlooks the role of cultivated attention, religious vocabulary, interpersonal expectations, and practices that teach participants how to identify and interpret unusual subjective events. Luhrmann et al. (2010) Van Leeuwen & van Elk (2019)

Luhrmann, Nusbaum, and Thisted studied how evangelical Christian practices and absorption-related differences relate to learning to experience communication from God. Their anthropological and psychological evidence suggests that socially instructed imagination and attention shape the salience and experiential vividness of divine interaction for participants. The broader inference is that religious cognition is enacted through practices that recruit ordinary mental capacities in culturally specific ways. Such findings do not determine the metaphysical origin of a reported event; they clarify the cognitive and cultural conditions under which particular experiences become available and interpretable. Luhrmann et al. (2010)

Cross-cultural work on models of mind complicates assumptions about a universal boundary between an internal thought and a message from an external spiritual source. In some settings, mental experiences are interpreted through a relatively bounded model of mind; in others, a more porous understanding permits thought, sensation, and influence to cross a personal boundary. Weisman and Luhrmann examined such models in a large cross-cultural sample, inviting a more precise account of how expectations about what minds are can influence reports of spiritual presence. A participant's response to the same phenomenological interview may depend on local concepts of thought, presence, and agency. Weisman & Luhrmann (2025)

The Interactive Religious Experience Model treats expectations, perceptual input, active interpretation, and social learning as dynamically interdependent. This approach predicts that religious cues may alter how ambiguous experiences are noticed and classified, and that interpretation can reinforce the future accessibility of particular experiences. The model is an explanatory proposal that needs explicit manipulations, time-resolved experience sampling, and disconfirmation criteria. A careful study would compare people with different training histories while measuring sensory ambiguity, attentional state, interpretive prompts, and later event attribution. Without this separation, the same observed report may be explained by multiple mechanisms. Van Leeuwen & van Elk (2019) van Elk (2022)

Spiritual-presence research also demands sensitive distinctions between vivid imaginative engagement, ordinary perception, culturally valued experience, and experiences accompanied by impairment or distress. A supernatural interpretation alone is not a psychological diagnosis; conversely, significant distress or functional disruption merits assessment regardless of a respondent's religious framework. The present research object concerns the cognitive architecture of interpretation, so clinical evaluation of extraordinary experiences belongs primarily in the neighboring clinical dossier. This disciplinary boundary supports rigorous description without pathologizing religious language or inferring clinical safety from cultural familiarity alone. Exline & Wilt (2023) Weisman & Luhrmann (2025)

Analytic Thinking, Cognitive Reflection, and Belief Reports

Dual-process and related cognitive models distinguish relatively rapid, context-sensitive forms of inference from deliberate evaluation of reasons and alternatives. This distinction is often applied to religion through a hypothesis that intuitive reasoning makes supernatural concepts accessible whereas reflective reasoning can scrutinize their plausibility. Such a model is useful only when the measured constructs are specified: cognitive reflection performance, propensity to engage analytical thought, accuracy on reasoning tasks, momentary cognitive load, and endorsement of religious claims are different variables. No single task provides a complete measure of an individual's rationality, and a difference in average scores is not by itself a psychological explanation of belief. Gervais & Norenzayan (2012) van Mulukom et al. (2024)

Gervais and Norenzayan's 2012 work reported experiments in which cues intended to promote analytical thinking lowered self-reported religious belief. Its historical importance lies in moving an association into an experimental claim with a manipulable antecedent. To evaluate that claim, however, one must ask whether the prime changes the proposed cognitive process, whether demand characteristics explain the observed answer, and whether the measured outcome is stable religious commitment or an immediate report susceptible to framing. These questions became central when later researchers directly replicated a specific manipulation under stronger preregistration and multisite sampling conditions. Gervais & Norenzayan (2012) Sanchez et al. (2017)

The preregistered replication by Sanchez, Sundermeier, Gray, and Calin-Jageman involved 941 participants across multiple sites and reported d = 0.07 with a 95% confidence interval from −0.12 to 0.25, rather than a reduction in belief. The estimate is consistent with a negligible effect of that intervention and indicates that the initial experimental priming result was less reliable than its early visibility suggested. This counterevidence is especially consequential because claims of a causal route from transient analytic cognition to religious disbelief require manipulation checks and independent replication. Its scope remains the tested prime and outcome, leaving open other mechanisms by which deliberation may change explanations. Sanchez et al. (2017)

Cross-national observational work offers a different level of evidence. Stagnaro and Pennycook analyzed cross-sectional reports from 50,827 persons across 16 countries and found associations between analytic thinking and retrospectively reported religious belief change, including changes in different directions. Their design enlarges the social scope of the question and complicates a unidirectional theory that reflection necessarily reduces religiosity. But a retrospective account of having changed beliefs is not a prospectively observed transition, and cross-sectional associations cannot establish that analytic thinking produced a change. Their result is most appropriately treated as evidence of a relationship worth investigating through longitudinal and intervention designs. Stagnaro & Pennycook (2025)

Metacognitive trust also matters: people vary in how much they value intuition, deliberation, personal experience, and authority as routes to knowledge. Van Mulukom and colleagues examined relations between confidence in thinking styles and supernatural beliefs, indicating that the personal meaning of intuition or reflection may be theoretically relevant alongside performance measures. Future experiments should manipulate opportunities for reflection while measuring metacognitive preferences, doctrinal familiarity, and concrete attributions about the same event. Without this design, researchers may incorrectly interpret differences in religious affiliation, expressed identity, or education as effects of a unitary reasoning style. van Mulukom et al. (2024) Stagnaro & Pennycook (2025)

Developmental and Educational Foundations of Religious Explanation

Children's supernatural agency concepts emerge within a developing capacity to reason about intentions, invisibility, physical causation, testimonial authority, and social rules. Research on purpose and design suggests that children can favor teleological interpretations of natural phenomena, but the claim that they are intuitive theists is a theoretical interpretation of a cluster of developmental findings, not a diagnostic description of every child. The crucial developmental distinction is between having cognitive resources to imagine an extraordinary agent and acquiring the particular religious concepts endorsed by one's caregivers and community. Narrative, ritual, and testimony shape the content attributed to these capacities. Kelemen (2004) Richert & Corriveau (2022)

Religious instruction provides statements about what a sacred agent knows, can do, and values, but children also infer meanings from practices, observed adult commitment, and the treatment of extraordinary stories as factual or fictional. A caregiver's assertion about an invisible agent can be weighed differently from a teacher's explanation about microbes, depending on the child's understanding of evidence, trust, and social expectation. Direct study requires measures of the child's concept rather than solely family affiliation. Developmental comparisons must also account for vocabulary, story familiarity, executive demands, and the social risks of contradicting an adult interviewer. Richert & Corriveau (2022) Lanman & Buhrmester (2017)

An important conceptual progression concerns what children and adults regard as possible, likely, and actual. A child may know how to imagine a ghost without affirming its existence, just as an adult may understand a religious doctrine without being committed to it. Fictional representation, possibility judgment, and belief endorsement should be separately elicited. This distinction becomes especially important in cross-cultural narrative tasks, where the researcher may introduce invented beings that activate ordinary conceptual expectations without carrying the social authority of locally recognized sacred beings. The study of memory for such stimuli provides information about representational structure rather than about actual religious socialization. Boyer & Ramble (2001) Richert & Corriveau (2022)

Education can change explanatory resources while leaving room for coexistence and contextual choice. Natural science instruction supplies mechanistic causal models; cultural and familial transmission supplies moral, relational, and sacred interpretations. The student may apply each to different questions or integrate them into a hierarchy of explanations. Longitudinal classroom research should therefore measure whether children genuinely revise beliefs about causal processes, learn to identify explanatory levels, or merely adjust responses to perceived school norms. The evidence that multiple frameworks coexist across development provides grounds for such careful differentiation. Legare & Gelman (2008) Legare et al. (2012)

Cross-Cultural Generalization and Population Diversity

A theory claiming that religious cognition reflects general human architecture must be tested beyond a narrow selection of university students, Western religious categories, and English-language measures. Boyer and Ramble's cross-cultural recall experiments represent an important early attempt to assess a hypothesized cognitive constraint across settings; research on coexisting explanations in South Africa provides a different form of evidence concerning locally meaningful causal categories. These designs address different levels: one examines memory for experimenter-constructed narratives, and the other examines explanations of significant events in culturally situated life. Generality cannot be inferred by pooling them conceptually without preserving those distinctions. Boyer & Ramble (2001) Legare & Gelman (2008)

Cross-cultural research needs both comparability and descriptive adequacy. Translations of concepts such as God, spirit, curse, soul, fate, luck, or karma may create false equivalence when the traditions under comparison organize agency and causality differently. A survey phrased around a personal creator could underestimate supernatural explanation in a non-theistic setting; a question about witches may carry social and historical meanings beyond the researcher's intended latent construct. Qualitative elicitation, cognitive interviewing, local expert collaboration, and tests of measurement invariance can expose these issues before large statistical comparisons are interpreted. Exline & Wilt (2023) Richert & Corriveau (2022)

The social credibility of claims varies through household practice, institutional authority, and public display. The evidence that credibility-enhancing religious behavior predicts endorsement more consistently than some mentalizing measures places cultural learning within the explanation of belief-related cognition. Still, a predictor of endorsement does not identify the precise mechanism by which a speaker's behavior changes a listener's causal representation. Comparative studies can probe how explanations differ when the same event is described by an elder, a stranger, a medical professional, or a religious authority. Such manipulations may reveal when epistemic trust mediates interpretation and when existing sacred concepts provide a ready interpretive framework. Maij et al. (2017) Lanman & Buhrmester (2017)

Individual heterogeneity exists inside traditions. A person's belief intensity, formal theological training, personal history, exposure to religious practice, and current circumstances may shape both explicit doctrine and the agent model recruited for a particular task. A strong study samples believers, doubters, disaffiliated persons, and nonreligious individuals without treating any group as a homogeneous cognitive type. Multilevel analyses should distinguish population means from within-person variability and examine whether event features produce similar responses across communities. Interactions with education and reflective reasoning must be interpreted in light of the reliability and equivalence of the instruments used. Stagnaro & Pennycook (2025) Barrett & Keil (1996) Exline & Wilt (2023)

Models of mind extend the comparative frame beyond explicit religious affiliation. Weisman and Luhrmann reported associations in a U.S. sample of 1,779 adults between spiritual-presence experiences and the flexibility to employ more porous conceptions of mind with cultural support. This work concerns how the boundary between a mental event and an apparently external communication is interpreted. It does not provide evidence that all cultures have the same experience rates or that a specific unusual experience is generated by one necessary cognitive mechanism. Comparison requires locally valid conceptual elicitation and, where feasible, prospective observation of changes in expectations and experience. Weisman & Luhrmann (2025)

Moralizing Agents, Social Inference, and Evolutionary Hypotheses

Some representations of supernatural agents include a special capacity to know moral behavior, monitor intentions, or impose rewards and punishments. Such concepts may enter individual causal attribution when a misfortune is construed as punishment or a fortunate outcome as recognition of virtue. They can also influence judgments about accountability, reputation, and cooperation. This is a specific subtype of supernatural agency cognition; many religious traditions and everyday practices involve agents with other concerns. Researchers must distinguish concept content, inferred surveillance, behavior in a social game, and long-term community cooperation rather than merging them into an unrestricted claim that religious belief automatically produces morality. Atran & Norenzayan (2004) Purzycki et al. (2016) Exline & Wilt (2023)

Purzycki and colleagues examined beliefs in moralistic and punitive supernatural agents alongside allocation behavior in small-scale societies, contributing cross-cultural evidence to debates about supernatural monitoring and cooperation. The central inferential question is whether differences in allocation follow beliefs about supernatural observation, shared norms, social network structure, or other correlated features. Cross-sectional cultural patterns and laboratory-style games can illuminate specific relationships but cannot establish the historical origin of complex societies by themselves. The evidence contributes to an explanatory model linking sacred agent representations and social decisions, while the broader evolutionary thesis requires archaeological, historical, and comparative corroboration. Purzycki et al. (2016)

Evolutionary accounts of religious cognition often distinguish proximate psychological mechanisms from ultimate explanations concerning why cognitive capacities or cultural arrangements persist. A propensity to infer agency under ambiguity, a preference for striking narratives, and social attention to credible instructors are candidates at the proximate level. Whether these predispositions were selected for nonreligious functions, favored because of religious consequences, or amplified through cultural selection is a further question. Boyer's by-product explanation and Atran and Norenzayan's broader evolutionary synthesis introduced productive hypotheses, but evolutionary narratives become scientific explanations only when their assumptions yield discriminating predictions and confront alternative historical trajectories. Boyer (2003) Atran & Norenzayan (2004) van Elk (2022)

Religious agency cognition is also intertwined with affiliation, ritual, emotion, and political history, though these topics demand their own scientific analysis. The representation of a punitive deity may influence some moral expectations in a structured experiment yet be outweighed in everyday behavior by institutions, self-interest, relationships, or situational incentives. The psychologically salient question is when the representation becomes active and what it changes under otherwise comparable conditions. Relevant experiments should measure explicit belief in monitoring, accessibility of the concept, expectation of sanctions, social visibility, and behavior, with careful attention to potential demand effects. Purzycki et al. (2016) Atran & Norenzayan (2004)

Taken together, research on moralizing agents supports the inclusion of social incentives and normative frameworks within accounts of causal attribution. A misfortune may be interpreted through cosmic justice, while a moral violation may evoke an imagined observer and an expectation of punishment. Both processes rely on representations whose content is transmitted culturally. Understanding the mechanisms requires linking individual mental models to local social practice without assuming uniform effects across religions or converting a statistical association into a universal evolutionary account. Exline & Wilt (2023) Lanman & Buhrmester (2017) Purzycki et al. (2016)

Contemporary Experimental Evidence: Conceptual Coherence and Teleological Language in 2026

Recent studies refine two methodological assumptions inherited from earlier cognitive science of religion. First, the memorability of an extraordinary concept may depend on whether its properties form a comprehensible causal and conceptual structure, not simply on the number of intuitive expectations it violates. Second, agreement with a sentence that contains teleological language may reflect the sentence's semantic form rather than a commitment to intentional creation. Both developments sharpen the distinctions between reported judgment, cognitive representation, and the mechanism hypothesized to produce it. Their relevance is direct: the present dossier treats concept recall and purpose attribution as experimentally measurable processes, so new findings that identify competing explanations must alter the weight assigned to older theories. Sommer et al. (2026) Tiililä et al. (2026)

Sommer, Musolino, and Hemmer's 2026 study of the memorability of supernatural concepts tested coherence as an alternative account of why concepts containing many intuitive-ontological violations are sometimes recalled poorly. One study compared the coherence of participant-generated natural and supernatural concepts; another compared memory for high- and low-coherence concepts in natural and supernatural categories. The reported results suggested that low coherence could explain memory disadvantages attributed to excessive counterintuitiveness. The authors also questioned the ecological validity of some highly contrived experimental concepts. These findings qualify a simplified inverted-U claim about violation count by identifying another property of representations that may travel with it. They do not erase earlier evidence of conceptual expectation effects, but require a factorial design capable of separating violation number, coherence, and familiarity. Sommer et al. (2026) Boyer & Ramble (2001)

The coherence hypothesis leads to specific prospective predictions. If memory difficulty follows the loss of a mutually consistent representation, a supernatural concept containing several violations should become easier to recall when its unusual properties form an intelligible, causally connected pattern. Conversely, adding independent and poorly integrated features may reduce recall even without changing the number of violations. Studies should match complexity, salience, explanatory richness, and prior experience and should estimate both false recalls and accurate retention. Real cultural concepts may exhibit higher coherence than arbitrary laboratory stimuli because stories, ritual use, and repeated interpretation supply relational structure over time. The value of the new evidence lies in making this alternative tractable, not in establishing one universal mechanism of cultural transmission. Sommer et al. (2026) Willard et al. (2023)

Tiililä, Schlottmann, and Bechlivanidis reported two experiments involving 896 participants that investigated whether wording shapes judgments about teleology in nature. Participants more readily endorsed statements framed with relational language such as an 'in order to' connective than formulations that more strongly implied purpose or intentional causal origins. Follow-up justifications and probes distinguished judgments about relations, functions, and purposive creation. The results demonstrate a substantial risk in interpreting agreement with an ambiguous teleological sentence as direct evidence of an underlying supernatural design belief. The manipulations establish a proximal effect of language and judgment framing under the tested tasks, whereas developmental origins or enduring beliefs require additional investigation. Tiililä et al. (2026)

The implication for supernatural causal-attribution research is operationally important. An item asserting that a phenomenon exists 'in order to' yield a benefit may elicit a true statement about its consequences rather than an affirmation that an agent intended its existence. Researchers should separately ask whether respondents perceive a beneficial relation, a function, a purposive origin, and a specific supernatural designer. The distinction connects the 2026 findings to earlier work on children's purpose reasoning and adults' interpretation of meaningful personal events. Because target questions differ, a high rate of endorsement in one format cannot be used as an unambiguous measure of the others. Tiililä et al. (2026) Kelemen (2004) Banerjee & Bloom (2014)

Together, these new experimental contributions support a methodological principle underlying this monograph: a theory of religious cognition should identify the representational variable that actually explains the observed outcome. Counting conceptual violations without measuring coherence risks a misleading account of memory; counting teleological endorsements without testing semantic interpretation risks a misleading account of supernatural causation. Independent replications, preregistered comparisons of rival mechanisms, and culturally grounded stimuli can determine whether these refinements generalize to religious concepts used outside laboratory settings. Sommer et al. (2026) Tiililä et al. (2026)

Experimental Contradictions, Replication, and Model Comparison

The main empirical disputes concern whether religious representations emerge principally from automatic agent detection, general mentalizing, distinctive memory, reflective reasoning, cultural testimony, or interactions among these processes. A sound comparison does not select one favorable result per theory. It asks what each mechanism predicts under interventions capable of separating it from rivals. An agency-detection model predicts changes in agent reports under controlled uncertainty or threat; a mind-perception model predicts effects of mental-state attribution; a memory account predicts differential recall under matched materials; a social-learning model predicts sensitivity to source credibility; an attribution model predicts event-dependent choices among explanations. The theories partly address different outcomes, so apparent conflict may reflect differing explananda. Boyer & Ramble (2001) Maij et al. (2017) Van Leeuwen & van Elk (2019)

The discrepancy between the initial analytic-thinking experiments and their registered replication demonstrates a frequent source of overgeneralization: immediate task performance is treated as a durable change in an identity-relevant belief. In the replication, the observed confidence interval crossed zero and excluded a reliably large decrease of the originally expected sort. A researcher should report that result beside the initial study and identify possible moderators rather than preserving the earlier claim through vague references to cognitive style. Cross-sectional evidence about analytic thinking and retrospective religious change provides a different outcome and therefore cannot independently repair a failed causal manipulation. Gervais & Norenzayan (2012) Sanchez et al. (2017) Stagnaro & Pennycook (2025)

The agency-detection literature contains an equally important measurement lesson. Van Elk's biological-motion study found an association involving paranormal belief that was contingent on stimulus ambiguity and was unrelated to some conventional religious belief measures. Subsequent preregistered immersive testing did not confirm an expected simple effect of threat on detection. These studies differ in sampling, technology, manipulation, and outcome, but together warn against treating any detection bias as the direct root of theistic commitment. A stronger explanatory test would independently measure prior religious expectations, paranormal endorsement, perceptual response criterion, sensitivity, and later causal attribution. van Elk (2013) Szymanek et al. (2025)

Evidence from theory-of-mind research also requires comparison across distinct measurement levels. A capacity for mentalizing may be necessary to understand a representation of an intentional god, while observed differences in mentalizing ability may fail to predict who believes that such a god exists. The results from a large multi-study comparison favor taking credibility-enhancing displays and cultural transmission seriously. This is consistent with conceptual dependence on social cognition and weak predictive specificity of a particular cognitive test. It should motivate mechanistic mediation experiments rather than a conclusion that the function of theory of mind has been disproved. Norenzayan et al. (2012) Maij et al. (2017) Gervais (2013)

A model comparison should report fit and prediction on held-out data where feasible, specify which parameters represent stable differences and which vary with context, and test whether effects generalize across translated instruments. Multiple mechanisms may produce observationally equivalent distributions of answers. Strong evidence emerges when a proposed variable is manipulated, its hypothesized intermediate process is measured, the predicted change occurs in an independent sample, and rival pathways are constrained by the design. In religious cognition, these standards are particularly important because narratives, emotions, and social context can change together. The present evidence favors an interactive plural-mechanism description while leaving causal weights and cultural boundary conditions open to further study. Van Leeuwen & van Elk (2019) Willard et al. (2023)

Critical Scientific Synthesis and Research Limits

The best-supported finding is that religious and supernatural ideas are psychologically representable through ordinary cognitive resources, including conceptual categorization, memory, social inference, goal attribution, and causal reasoning. Experiments demonstrate that the form of a response depends on task instructions and representational demands. Counterintuitive properties can shape recall, explicit theological descriptions can diverge from story-processing inferences, and supernatural explanations can coexist with natural causal accounts. These results document processes and regularities without establishing a single general cause of religious belief or the truth status of the concepts represented. Barrett & Keil (1996) Boyer & Ramble (2001) Legare & Gelman (2008)

Several explanations remain plausible but limited in their generality. Hypersensitive agency detection offers a hypothesis about false-positive perception; evidence linking it to conventional religiosity is uneven. Mentalizing supports understanding of persons and intentional sacred agents, but individual performance on mentalizing tasks has limited demonstrated power to separate believers from nonbelievers. Analytic-thinking accounts require distinguishing observational relations from the causal effects of transient primes. Cultural credibility and explanatory context are predictive in some studies but remain subject to confounding and measurement challenges. The integrative conclusion is that mechanism, representation, endorsement, and event-level attribution should be analyzed as separate coupled processes. van Elk (2013) Maij et al. (2017) Sanchez et al. (2017) Lanman & Buhrmester (2017)

The field has persistent sampling constraints. Some foundational experiments recruited predominantly educated adults from limited cultural settings, while wider comparative studies often rely on survey measures whose meanings vary across communities. A study conducted in one tradition may use the word God as though it were an interchangeable label for all supernatural concepts. A cross-cultural task may impose an artificial stimulus unfamiliar to participants, strengthening internal stimulus control at the cost of relevance to a real religious community. Replication across countries helps only when the construct and its measurements are equivalent. Researchers need jointly standardized and locally grounded methods. Boyer & Ramble (2001) Richert & Corriveau (2022) Exline & Wilt (2023)

Many proposed mechanisms are susceptible to alternative explanations. Reported presence may depend on imagination, sensory ambiguity, attention, and interpretive training; moralizing-agent effects may reflect reputational concerns and local norms; teleological judgments may express meaning or design depending on the question; analytic response scores may covary with education and social identity. The practical methodological response is to compare accounts within a common task rather than accumulating isolated correlations across incomparable outcomes. Prospective and experimental designs can constrain temporal order, and qualitative research can establish what participants mean by their causal language. Luhrmann et al. (2010) Purzycki et al. (2016) Banerjee & Bloom (2014) Stagnaro & Pennycook (2025)

Scientific synthesis must also recognize the boundary of laboratory access. An experiment can test whether a visual stimulus elicits reports of agency, whether a story is recalled with humanlike distortions, or whether event descriptions alter supernatural explanations. It does not adjudicate metaphysical claims about gods, spirits, karma, or fate. Respecting this distinction strengthens, rather than weakens, the research: testable psychological propositions remain identifiable, and participants' traditions can be described faithfully without equating reporting with error or cultural significance with proof. Researchers should clearly identify when their language describes a participant's attribution rather than an independently verified external cause. Exline & Wilt (2023) Weisman & Luhrmann (2025)

Future Research: Integrating Mechanisms Across Context and Time

A high-value experimental program would cross agent visibility, event ambiguity, event valence, and culturally relevant causal prompts in a preregistered design. Participants would first make perceptual classifications, then report whether an intentional actor was involved, then assign possible natural and supernatural causes, and finally describe the meaning of the event. Such sequencing would reveal whether an agency judgment precedes a supernatural inference, whether both derive independently from expectations, and when a nonpersonal sacred explanation is preferred. Signal detection analysis could quantify sensitivity and response criterion, while open-ended responses would capture categories missing from fixed-choice instruments. van Elk (2013) Ramsay et al. (2023) Exline & Wilt (2023)

Longitudinal studies should follow the same individuals across instruction, participation in practices, major life events, and changes in their social networks. Repeated measures of explicit agent concepts, narrative inference, supernatural attribution, reasoning preference, and lived experience could distinguish stable tendencies from momentary contexts. The study of belief acquisition and revision requires a separate causal and developmental emphasis, but a cognitive monograph can contribute precise repeated measures of the representations being acquired or changed. Any model of change must estimate attrition, selection, and alternative temporal pathways, particularly when changes coincide with schooling, migration, loss, or shifts in community membership. Richert & Corriveau (2022) Stagnaro & Pennycook (2025)

Cross-cultural projects should combine local ethnographic concept elicitation with shared experimental logic. Researchers can first map indigenous terms for spiritual agents and causation, then develop equivalent tasks that do not impose a single religion's ontology. Multiple communities could test whether comparable effects depend on counterintuitiveness, ambiguity, social credibility, or specific religious instruction. Successful replication would be demonstrated by consistent operational contrasts and interpretable effect heterogeneity, not by forcing all populations to produce identical scores on a culturally narrow instrument. Normative judgments about the plausibility of a tradition should remain separate from measurement quality. Boyer & Ramble (2001) Maij et al. (2017) Legare et al. (2012)

Computational models offer a way to make interaction theories precise. A probabilistic observer could combine prior beliefs about agency, the reliability of sensory evidence, the anticipated costs of a missed actor, and testimony from trusted informants. An explanatory-choice component could map the posterior assessment of an event into available categories such as coincidence, human intent, divine providence, or fate. This is a proposed modeling strategy rather than a finding about the actual algorithm used by religious thinkers. Its value would depend on prospective quantitative prediction and on experiments manipulating identifiable priors, evidence, and source credibility. Van Leeuwen & van Elk (2019) van Elk (2022)

Studies of spiritual experience can combine experience sampling with experimental cues and transparent measures of absorption and models of mind. These methods could address whether momentary reports of spiritual presence follow changes in attentional state, perceived permeability of the mind, social context, or previous interpretation. The causal question is best framed at an observable level: which factors predict the timing and description of a reported experience? Deliberate attention to distress and individual autonomy should accompany participant recruitment and interpretation. Luhrmann et al. (2010) Weisman & Luhrmann (2025)

Interdisciplinary Scope and the Artificial Era

Religious cognition connects developmental psychology, cognitive science, cultural psychology, social inference, anthropology, neuroscience, and the philosophy of causal explanation. Each domain contributes a distinctive level of analysis. Developmental work specifies changes in conceptual resources and testimony; cognitive experiments isolate mechanisms of memory or judgment; cultural research identifies local categories and practices; social psychology examines credibility and group expectations; anthropology documents how extraordinary agency is situated in lived environments. A comprehensive account needs these levels to be linked through explicit hypotheses rather than merged through a single loose category of religion. Richert & Corriveau (2022) Maij et al. (2017) Exline & Wilt (2023)

Within the Artificial Era, research on human attribution of mind and agency to interactive artificial systems offers a limited but informative comparison. Humans may assign intentionality or communicative presence to a conversational system partly through processes of anthropomorphism and social inference. Religious agent representations, however, carry theological and cosmological commitments, cultural histories, ritual obligations, and causal roles that cannot be inferred merely from an interface eliciting social language. The appropriate research question is which cognitive principles of agent representation apply across these domains, and which are specific to religious causal expectations. A functional analogy should be assessed with comparable measures rather than assumed on philosophical grounds. Epley et al. (2007) Gervais (2013) Van Leeuwen & van Elk (2019)

This comparison preserves the publication's scientific boundary. Human attribution to artificial systems is a psychological report about the observer and interaction; it cannot by itself establish the internal psychology, consciousness, or metaphysical status of the system involved. The broader philosophical framework of Artificial Era and From Homo to Artificial identifies an intellectual horizon in which such comparisons are worth pursuing. Claims concerning the ontology of Artificial are an attributed philosophical project, whereas the empirical propositions developed here concern measurable human cognition, explanatory judgment, and cultural transmission. Separate dossiers on human–artificial attribution and psychological organization can study those questions in depth, provided they are linked only when actually published.

Transdisciplinary research would profit from parallel tasks asking participants to infer the knowledge, goals, physical capacities, and accountability of an ordinary person, a named supernatural figure, a fictional extraordinary agent, and an artificial interlocutor. The manipulation would distinguish cultural familiarity, perceived autonomy, and violated conceptual expectations while holding the wording of relevant mental-state questions as comparable as possible. The experiment would not imply equivalent ontological status among targets; it would measure how human respondents allocate agent properties across conceptual classes. Such a design directly extends the methodological lessons of narrative comprehension, mind perception, and anthropomorphism research. Barrett & Keil (1996) Epley et al. (2007) Gervais (2013)

Conclusion

Religious cognition organizes extraordinary agency and causal explanation through interacting capacities for conceptual representation, memory, social understanding, purpose attribution, uncertainty management, and culturally guided interpretation. Experimental studies establish that explicit theological claims may diverge from intuitive narrative inferences, that limited counterintuitive violations can affect recall, and that supernatural causal accounts can coexist with biological or social explanations. These findings define a researchable architecture of processes whose activation depends on the representation, task, personal history, and social setting. Barrett & Keil (1996) Boyer & Ramble (2001) Legare & Gelman (2008)

The critical scientific record favors specificity. A perceptual bias toward illusory agents has been demonstrated in some paranormal-belief comparisons, yet that result does not generalize automatically to institutional religiosity. A causal claim about analytic-thinking primes was substantially weakened by a preregistered replication. A simple account from mentalizing capacity to religious endorsement has encountered extensive cross-cultural counterevidence. Cultural credibility, learned expectations, event characteristics, and models of mind help explain why the same cognitive resources support diverse interpretations. Each mechanism must be tested with its own observable predictions. van Elk (2013) Sanchez et al. (2017) Maij et al. (2017) Exline & Wilt (2023) Weisman & Luhrmann (2025)

The resulting synthesis treats supernatural-agent concepts, causal attribution, personal endorsement, and religious identity as related but distinguishable scientific objects. For the research of religious cognition, the central challenge is to identify when a person represents an agent, when they believe such an agent exists, when they assign it causation, and how these judgments change across context and time. This framework guides future comparative experiments, longitudinal studies, culturally valid measurement, and interdisciplinary modeling. Its explanatory power will be determined by independent replication, transparent measures, and the capacity to discriminate between rival models rather than by the breadth of its philosophical appeal. Van Leeuwen & van Elk (2019) Richert & Corriveau (2022) Willard et al. (2023)

References

  • Atran, S., & Norenzayan, A. (2004). Religion's evolutionary landscape: Counterintuition, commitment, compassion, communion. Behavioral and Brain Sciences, 27(6), 713–730. https://doi.org/10.1017/S0140525X04000172
  • Banerjee, K., & Bloom, P. (2014). Why did this happen to me? Religious believers' and non-believers' teleological reasoning about life events. Cognition, 133(1), 277–303. https://doi.org/10.1016/j.cognition.2014.06.017
  • Barrett, J. L., & Keil, F. C. (1996). Conceptualizing a nonnatural entity: Anthropomorphism in God concepts. Cognitive Psychology, 31(3), 219–247. https://doi.org/10.1006/cogp.1996.0017
  • Boyer, P. (2003). Religious thought and behaviour as by-products of brain function. Trends in Cognitive Sciences, 7(3), 119–124. https://doi.org/10.1016/S1364-6613(03)00031-7
  • Boyer, P., & Ramble, C. (2001). Cognitive templates for religious concepts: Cross-cultural evidence for recall of counter-intuitive representations. Cognitive Science, 25(4), 535–564. https://doi.org/10.1016/S0364-0213(01)00045-3
  • Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. https://doi.org/10.1037/0033-295X.114.4.864
  • Exline, J. J., & Wilt, J. A. (2023). Supernatural attributions: Seeing God, the devil, demons, spirits, fate, and karma as causes of events. Annual Review of Clinical Psychology, 19, 461–487. https://doi.org/10.1146/annurev-clinpsy-080921-081114
  • Gervais, W. M., & Norenzayan, A. (2012). Analytic thinking promotes religious disbelief. Science, 336(6080), 493–496. https://doi.org/10.1126/science.1215647
  • Gervais, W. M. (2013). Perceiving minds and gods: How mind perception enables, constrains, and is triggered by belief in gods. Perspectives on Psychological Science, 8(4), 380–394. https://doi.org/10.1177/1745691613489836
  • Guthrie, S. E. (1993). Faces in the clouds: A new theory of religion. Oxford University Press. https://doi.org/10.1093/oso/9780195069013.001.0001
  • Kelemen, D. (2004). Are children 'intuitive theists'? Reasoning about purpose and design in nature. Psychological Science, 15(5), 295–301. https://doi.org/10.1111/j.0956-7976.2004.00672.x
  • Lanman, J. A., & Buhrmester, M. D. (2017). Religious actions speak louder than words: Exposure to credibility-enhancing displays predicts theism. Religion, Brain & Behavior, 7(1), 3–16. https://doi.org/10.1080/2153599X.2015.1117011
  • Legare, C. H., & Gelman, S. A. (2008). Bewitchment, biology, or both: The co-existence of natural and supernatural explanatory frameworks across development. Cognitive Science, 32(4), 607–642. https://doi.org/10.1080/03640210802066766
  • Legare, C. H., Evans, E. M., Rosengren, K. S., & Harris, P. L. (2012). The coexistence of natural and supernatural explanations across cultures and development. Child Development, 83(3), 779–793. https://doi.org/10.1111/j.1467-8624.2012.01743.x
  • Luhrmann, T. M., Nusbaum, H., & Thisted, R. (2010). The absorption hypothesis: Learning to hear God in evangelical Christianity. American Anthropologist, 112(1), 66–78. https://doi.org/10.1111/j.1548-1433.2009.01197.x
  • Maij, D. L. R., van Harreveld, F., Gervais, W., Schrag, Y., Mohr, C., & van Elk, M. (2017). Mentalizing skills do not differentiate believers from non-believers, but credibility enhancing displays do. PLOS ONE, 12(8), Article e0182764. https://doi.org/10.1371/journal.pone.0182764
  • Norenzayan, A., Gervais, W. M., & Trzesniewski, K. H. (2012). Mentalizing deficits constrain belief in a personal God. PLOS ONE, 7(5), Article e36880. https://doi.org/10.1371/journal.pone.0036880
  • Purzycki, B. G., Apicella, C., Atkinson, Q. D., Cohen, E., McNamara, R. A., Willard, A. K., Xygalatas, D., Norenzayan, A., & Henrich, J. (2016). Moralistic gods, supernatural punishment and the expansion of human sociality. Nature, 530, 327–330. https://doi.org/10.1038/nature16980
  • Ramsay, J. E., Khong, Z. Y., & Yeo, J. S. E. (2023). The God of big (and good) things: Religious priming, event properties, and supernatural explanations. The International Journal for the Psychology of Religion, 33(2), 89–106. https://doi.org/10.1080/10508619.2022.2097778
  • Richert, R. A., & Corriveau, K. H. (2022). Development of religious cognition. Annual Review of Developmental Psychology, 4, 185–206. https://doi.org/10.1146/annurev-devpsych-120920-041303
  • Sanchez, C., Sundermeier, B., Gray, K., & Calin-Jageman, R. J. (2017). Direct replication of Gervais & Norenzayan (2012): No evidence that analytic thinking decreases religious belief. PLOS ONE, 12(2), Article e0172636. https://doi.org/10.1371/journal.pone.0172636
  • Sommer, J., Musolino, J., & Hemmer, P. (2026). Coherence and the memorability of supernatural concepts. Cognition, 276, Article 106661. https://doi.org/10.1016/j.cognition.2026.106661
  • Stagnaro, M. N., & Pennycook, G. (2025). On the role of analytic thinking in religious belief change: Evidence from over 50,000 participants in 16 countries. Cognition, 254, Article 105989. https://doi.org/10.1016/j.cognition.2024.105989
  • Szymanek, P., Nenadalová, J., & Van Leeuwen, N. (2025). Revisiting feeling of threat and agency detection: A preregistered virtual reality study. Psychology of Religion and Spirituality, 17(3), 161–174. https://doi.org/10.1037/rel0000538
  • Tiililä, T., Schlottmann, A., & Bechlivanidis, C. (2026). Teleological language or teleological thinking? Cognition, 274, Article 106551. https://doi.org/10.1016/j.cognition.2026.106551
  • van Elk, M. (2013). Paranormal believers are more prone to illusory agency detection than skeptics. Consciousness and Cognition, 22(3), 1041–1046. https://doi.org/10.1016/j.concog.2013.07.004
  • van Elk, M. (2022). Proximate and ultimate causes of supernatural beliefs. Frontiers in Psychology, 13, Article 949131. https://doi.org/10.3389/fpsyg.2022.949131
  • Van Leeuwen, N., & van Elk, M. (2019). Seeking the supernatural: The Interactive Religious Experience Model. Religion, Brain & Behavior, 9(3), 221–251. https://doi.org/10.1080/2153599X.2018.1453529
  • van Mulukom, V., Baimel, A., Maraldi, E., & Farias, M. (2024). Examining the relationship between metacognitive trust in thinking styles and supernatural beliefs. Scandinavian Journal of Psychology, 65(2), 206–222. https://doi.org/10.1111/sjop.12961
  • Weisman, K., & Luhrmann, T. M. (2025). Shifting between models of mind: New insights into how human minds give rise to experiences of spiritual presence and alternative realities. Topics in Cognitive Science, 17(2), 144–179. https://doi.org/10.1111/tops.70002
  • Willard, A. K., Turpin, H., & Baimel, A. (2023). Universal cognitive biases as the basis for supernatural beliefs: Evidence and critiques. In J. J. Tehrani, J. Kendal, & R. L. Kendal (Eds.), The Oxford handbook of cultural evolution (pp. 567–578). Oxford University Press. https://doi.org/10.1093/oxfordhb/9780198869252.013.46