Mysterious illnesses have supernatural and ritualistic cures: Evidence from 3,655 century-old Irish folk cures.

Mysterious illnesses have supernatural and ritualistic cures: Evidence from 3,655 century-old Irish folk cures - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice Proc Natl Acad Sci U S A . 2025 Dec 1;122(49):e2511006122. doi: 10.1073/pnas.2511006122 Mysterious illnesses have supernatural and ritualistic cures: Evidence from 3,655 century-old Irish folk cures Mícheál de Barra a, 1 , Ángel V Jiménez a , Nachita Rosun a, b , Aiyana K Willard a Author information Article notes Copyright and License information a Centre for Culture and Evolution, Department of Psychology, Brunel University of London, Uxbridge UB8 3PH, United Kingdom b Centre for Psychological Research, Oxford Brookes University, Oxford OX3 0BP, United Kingdom 1 To whom correspondence may be addressed. Email: micheal.debarra@brunel.ac.uk . Edited by Joyce Marcus, University of Michigan, Ann Arbor, MI; received May 9, 2025; accepted October 13, 2025 Received 2025 May 9; Accepted 2025 Oct 13; Issue date 2025 Dec 9. Copyright © 2025 the Author(s). Published by PNAS. This article is distributed under Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND) . PMC Copyright notice PMCID: PMC12704799  PMID: 41325525 See commentary " Explaining supernatural cures for poorly understood illnesses: Revisiting classic questions with new evidence " in volume 123, e2531788123. Significance Classical anthropological and cognitive theories propose that supernatural healing practices emerge when ordinary causal reasoning fails, yet direct quantitative tests remain scarce. Using 3,655 “local cures,” collected as part of a national project to document folklore in Ireland in 1937–1938, we quantitatively test a range of theories about the appeal of supernatural cures. Preregistered mixed-effects models reveal that diseases whose causes or bodily mechanisms would have eluded lay observers were around 50% more likely to attract religious or magical treatments, whereas disease severity, pain, anxiety, and need for care showed no reliable relationship with supernatural or religious cure content. These findings suggest epistemic uncertainty may be a driver of supernatural thinking about health. Keywords: folk medicine, cultural evolution, cognitive science of religion, health psychology, ritual psychology Abstract Why and when do people draw upon religious and supernatural solutions to problems? Cognitive scientists and anthropologists have proposed a range of answers, stressing religion and ritual’s capacity to alleviate anxiety, create a sense of order, or explain otherwise inexplicable events. Here, we leverage a unique dataset of 3,655 folk cures for 35 diseases, collected in 1937/8 from a mostly rural Irish sample born roughly between 1850 and 1925. Since the diseases vary in theory-relevant ways and the cures vary in the degree to which they include religious and supernatural elements, this dataset facilitates a unique test of these predictions in a premodern western population. In preregistered tests, we find that diseases judged by two doctors to have causes and mechanisms that would be unclear to the patients were more likely to have supernatural/religious treatments. Contra common predictions, severe and disabling diseases did not have more supernatural/religious cures and anxiety-provoking diseases did not have more ritualistic cures. The causes and treatments of illness have been regularly linked to supernatural and religious beliefs across human cultures and throughout time ( 1 ). Illnesses everywhere are thought to be plausibly caused by spirits, the evil eye, and breaking social taboos, and believed to be treatable with prayers, amulets, rituals, and shamanic healing. These supernatural beliefs exist alongside more naturalistic frameworks emphasizing, for example, contamination, herbal remedies, as well as more biomedically informed concepts ( 2 ). Within any given cultural group, there is typically substantial cross-disease variability in the extent to which people’s explanatory models and responses draw upon supernatural vs. naturalistic worldviews. Epilepsy, for example, is very often seen in supernatural terms ( 3 ) whereas cuts and stings more rarely acquire supernatural explanations. This variation provides a useful opportunity to test the explanatory power of theories about the function and role of supernatural and religious beliefs since different theories make distinct predictions about which kinds of contexts—and hence which diseases—are most likely to elicit supernatural responses. Careful comparison of supernatural and naturalistic responses to a range of different diseases may help researchers understand the role of religion in this important area—they reveal how people think and make decisions about illness, as well as shedding light on when we are prone to use supernatural and religious explanations more generally. Why might people draw upon supernatural or religious concepts, rather than naturalistic ones, when attempting to explain and remedy ill health? Theories within the cognitive, anthropological, and evolutionary literatures can be used to derive several testable hypotheses, premised on the concept that treatments will spread more readily if they satisfy some psychological or social need. In brief, this literature suggests that 1) diseases with no obvious cause or mechanism (i.e., higher causal opacity or uncertainty) should be more likely to be treated with supernatural or religious cures 2) diseases that are more severe—i.e., that have a higher case fatality rate—should be more likely to elicit religious treatments, 3) diseases that provoke anxiety should be more likely to elicit ritualistic treatments and 4) diseases that incapacitate the patient and, therefore, increase the need for care and support, should elicit more religious treatments. Though we acknowledge these factors are interconnected (e.g., uncertainty can increase anxiety, severity can increase uncertainty), we will discuss them separately to differentiate the theoretical basis of each. Despite the prominence of these theories, little work has examined how different types of existential threats influence the expression of religious/spiritual belief and behavior within a single domain such as illness, and even fewer have done so in a way that allows for comparison with the use of scientific/naturalistic explanations under similar circumstances. By looking at factors like uncertainty, severity, anxiety, and need for support together, we can differentiate between them, to draw more precise conclusions about if, when, and why different meaning-making systems are engaged. Natural, Supernatural, and Religious. How ought we distinguish between natural, supernatural, and religious treatments? “Supernatural” can be defined as things beyond scientifically verifiable natural laws, but this definition does little to help us understand how people for much of human history might have thought about this category. For the 1930s Irish people who generated our dataset, “verifiable and natural” are more likely to consist of observable causal processes like physical touch or consuming a substance than scientific principles. Supernatural beliefs that are “beyond” natural consist of more complex causal processes, like actions at a distance, or the ideas of similarity in sympathetic magic. Though people do seem to be able to think about both these categories as distinct and sometimes co-occurring causes of illness ( 4 , 5 ), it is not straightforward to specify the exact characteristics that consistently differentiated one from the other. The belief that a disease is caused by an invisible spirit is in many ways like the belief that disease is caused by invisible bacteria, and researchers have emphasized supernatural and natural thinking draws on some similar patterns of cognitive processing ( 6 , 7 ). Identifying what counts as “religious” is not straightforward either. For our purposes, we are interested in religious beliefs as they appear in folk treatments. Although religious institutions encompass more than the supernatural, religious folk beliefs can be regarded as a subset of supernatural beliefs. Here, we focus on practices such as the use of holy water or religious sites within folk treatments—these represent a specific subset of the complex causal processes we classify as supernatural. We recognize that people using these cures might not sharply distinguish natural vs. supernatural. Nevertheless, this distinction is useful for testing theories of the origins of supernatural concepts outlined below and provides a basis for drawing a line to separate two categories that likely blur together somewhat in most people’s minds. Diseases with No Obvious Cause or Mechanism. Malinowski ( 8 ) argued that supernatural reasoning is more readily applied where the causal processes that determine outcomes are not observable, comprehensible, or obviously modifiable. For instance, people understand perfectly well the labor and skill needed to build canoes or maintain a productive garden and that magic and ritual cannot substitute for these quite nonmagical investments. But they also recognise that good harvests and successful fishing expeditions are a function of other less tractable processes. Supernatural rituals are directed toward these latter unpredictable aspects. For example, deep sea fishing is more imbued with magic and ritual than low-risk fishing within the lagoon ( 8 ). More relevant to our current project, Malinowski noted that diseases with obvious explanations—obvious within a given medical framework —tended to elicit materialistic explanations (e.g., eating too much, working too hard) whereas diseases with less obvious explanations were attributed to witchcraft. A straightforward prediction from this perspective is that diseases which are less comprehensible will attract more magical and religious cures. This effect has been suggested as part of the origin of shamanistic practices ( 9 ), and more recently, similar effects of uncertainty around the causes of illness have been found experimentally in Mauritius ( 10 ). When causes of mystery diseases were portrayed as unclear, participants saw supernatural causes as more likely and diseases with supernatural causes as requiring supernatural cures. But why might people find supernatural cures more compelling when the etiology (causes of an illness) and pathology (symptoms and progression) are unclear? Hong ( 11 ) and others ( 12 ) have argued that people find it psychologically difficult to accept ignorance and ineffectiveness when faced with opaque processes like, presumably, some illnesses. Working under assumptions of determinism (everything has a cause) and cosmological holism (all events are interconnected and these webs of connection span natural and supernatural domains), people seek ways of changing important outcomes despite the apparent intractability ( 13 ). Contrast tuberculosis (TB), with no obvious cause and unobservable disease processes, with a cut that has an obvious cause. A bandage or a salve does not seem like a plausible treatment for TB but does for a cut. Opaque illnesses like TB present fewer options for obvious naturalistic cures. When causes and processes of a disease are unclear, we might expect a greater reliance on more general-purpose supernatural cures. Cases where diseases are given both supernatural and natural causes show the same pattern ( 5 ). HIV can be believed to be caused by both a virus and witchcraft: the virus explains the disease process, while witchcraft explains why this particular person encountered the virus (see also ref. 14 ). These are not competing explanations for the same level of causation. The natural cause accounts for the biomedical mechanism; the supernatural cause addresses the more opaque social or moral “why-me” question ( 10 ). Religious cures may also help people recover from the destabilizing effects of illness by providing them with a sense that these events, while disruptive, are under some form of external control and influence ( 12 ). The sense that one’s illness is incoherent and unpredictable is associated with a reduced ability to cope and plan, as well as poorer survival, over and above the direct effects of the pathology ( 15 ). Cures framed in a religious/supernatural way may be particularly good in helping people gain a sense of control over their circumstances. If these theories accurately characterize people’s responses to illness, then we should see greater use of supernatural worldviews in response to illnesses with uncertain causes or which involve unfamiliar combinations of symptoms, relative to illnesses that are more coherent and predictable. Severity. Religious and supernatural cures might be particularly important for severe illnesses as these illnesses may provoke existential concerns. According to some scholars, the religious and spiritual ideas have the capacity to alleviate mortality anxiety and other types of existential concerns ( 16 , 17 ). Religious and supernatural beliefs that suppose an afterlife or nonmaterial spiritual world are argued to give solace to people whose material existence is threatened by illness and the reminders of death ( 17 , 18 ), and can offer comfort in times of upheaval and misfortune ( 19 ). Indeed, some studies have found that reminding people of their own mortality led to stronger religious beliefs and a stronger belief in divine intervention ( 20 ). More ecologically valid studies have found increases in religiosity among terminally ill people ( 21 ) and people who experienced hardship due to a natural disaster ( 19 ). However, other studies have seen little effect ( 10 , 22 ) and so there is a clear need for additional tests of the explanatory power of these theories. Ritual and Anxiety. Anxiety has clear adaptive benefits in situations of potential danger when one’s actions can have an immediate impact on the anxiety-causing stimuli such as fighting or fleeing a predator ( 23 ). When this is not the case, anxiety can become chronic and damaging. Repetitive and ritualistic behaviors may provide relief from such maladaptive anxiety. Why? At a cognitive level, ritual actions are argued to provide a period of high-prediction accuracy, enabling the person to recover a sense of control ( 24 ) and thus improving performance, decision-making, and well-being ( 25 ). The capacity of ritualized behavior to provide relief from anxiety may have deep evolutionary roots ( 26 ) with nonhuman animals also displaying repetitive ritualistic-like behaviors such as pacing in conditions of chronic anxiety ( 27 ). In humans, the stress that comes from the ability to ruminate on existential questions provides additional scope for chronic anxiety, especially during periods of illness. According to theories of ritual, anxiety-provoking illnesses should therefore have more ritualistic components in their treatments (e.g., performing a set of steps several times unnecessarily) than more benign illnesses. Anxiety is likely one of the underlying mechanisms of the effects of severity and uncertainty and can be present even in illnesses for which we see a tractable cause. Thus, the use of ritual should be prevalent across both supernatural/religious and naturalistic treatments. Religious Care. Some evolutionary theories suggest that religion has the adaptive benefits of helping humans cooperate in demanding situations ( 28 , 29 ), solidifying social bonds, and redistributing material resources in the face of adversity ( 30 , 31 ). Believers of some gods consider that gods reward people who carry out acts of charity and care for the sick. Theorists like Stark ( 32 ) have argued that part of what made early Christianity spread was the mandate to look after the sick and needy. The higher survival rate of Christians during plagues, and general tendency to do better through their cooperative efforts, was noticed by others, seen as evidence of the power of their God, and caused more people to convert. Illness and injury can cause long periods of incapacity where the victim is unusually vulnerable and dependent on care from kith and kin. For example, Sugiyama ( 33 ) found that 90% of Shiwiar forager-horticulturalists spent a month or more incapacitated by illness or injury in their lives; without extensive social support, such a period of incapacity will be fatal. Rather than the treatments themselves helping people who are ill, the adaptive benefit may lie in their ability to encourage social support. They do so by signaling that the sufferer is a good cooperative partner and worthy of care by others in the community ( 34 ) to increase one’s chance of surviving an illness (e.g., ref. 35 ). The onset of illness may thus be a prudential moment to demonstrate a commitment to the relevant religious community via engaging in a faith-based treatment ( 35 , 36 ). From this material benefits theory, we can suggest that religious treatments (as a subset of the broader category of supernatural treatments) will be more common when illness requires more care. The Schools’ Collection from the Irish National Folklore Collection. We examine supernatural/religious and naturalistic content in a large database of folk medical cures for a broad range of diseases collected in the 1930s in Ireland, when supernatural and natural folk cures were in wide usage ( 37 ). This digitalized dataset includes over 740,000 pages of folklore, including folk cures collected across the Irish Free State, soon after its formation. The cures were collected as part of a broader folklore project involving 50,000 children and 5,000 schools ( 38 ). The children interviewed parents, grandparents, and elders on local history, traditions, farming practices, as well as folk medical practices. Their records were transcribed and collected under the supervision of teachers to create a unique dataset which may illuminate, among other things, medical and religious culture in a sample of people born between 1850 (the likely birth year of the oldest informants) and 1925 (the data-collecting children). The cures include a range of naturalistic cures (e.g., to cure a nosebleed, draw blood from the face by applying a cold object to the back of the neck), religious cures (e.g., holy water is poured into an aching ear while saying a prayer), and other supernatural cures (e.g., put a worm in a bottle and bury it where it cannot be found to cure a toothache). Since the diseases that these cures treat vary along theoretically relevant dimensions like seriousness and causal opacity, they provide a unique testing ground for theories which predict when religious and supernatural cures will dominate over more naturalistic treatments. One might expect naturalistic treatments to outcompete supernatural and religious cures because they are more effective at alleviating symptoms or shortening illness. However, cross-culturally and historically, folk cures often rely on intuitive but implausible biological processes and, with some exceptions, are unlikely to benefit people directly ( 39 , 40 ). A cursory examination of the treatments in the corpus suggests that the vast majority have no direct beneficial effects. Hypotheses. Based on the theories outlined above, we have formulated the following hypotheses. We were largely agnostic on which of these hypotheses would be supported but used them to structure our search for patterns in the uses of supernatural and naturalistic cures. H1: Illnesses with unclear causes and unclear underlying pathology will be more commonly treated with supernatural treatments (including religious treatments). H2: Severe illnesses will be more commonly treated with supernatural treatments (including religious treatments) than with naturalistic treatments. H3: Ritualistic treatments (both natural and religious/supernatural) will be more common when illnesses provoke high anxiety. H4: Explicitly religious treatments (as opposed to other types of supernatural content) will be more common when illness results in a need for care. In addition to these preregistered hypotheses ( https://osf.io/ed8v9 ), we performed some exploratory analyses. All confirmatory and exploratory tests are clearly labeled below. The dataset and the analysis code are also available on figshare ( https://figshare.com/s/77c9c858581e5014feed ). Results Descriptive Statistics. The most frequent diseases in the dataset were warts (N = 576), followed by whooping cough (N = 365), toothache (N = 225), and burns (N = 190). The diseases with the greatest proportion of supernatural nonreligious treatment were whooping cough (72%), mumps (66%), and warts (53%). Religious treatments were represented in a greater proportion in treatments for sties (40%), erysipelas (28%), and rickets (25%). Corns (100%), cuts and wounds (95%), and earache (94%) were treated with naturalistic treatments more than other diseases. Mumps (74%), sties (63%), and rickets (48%) were most often ritualistic while cuts and wounds (100%), broken bones (98%), and bleeding (97%) were most often not ritualistic. The tables in part 1 of SI Appendix show the descriptive statistics for the predictor and outcome variables for each disease, respectively. Uncertainty About the Causes and Mechanisms of Illness. We predicted (H1) that treatments would be more likely to have supernatural and religious content if the illnesses had no clearly identifiable cause or mechanism. To test this, we used two proxies for uncertainty based on the ratings provided by two medical doctors: uncertainty about the cause and uncertainty about the mechanism. According to our medical experts, the antecedent causes of illnesses like rheumatism (rating = 9), rickets (rating = 9), and cancer (rating = 8.5) would be unclear to people without a medical education. In contrast, sprains, stings, and burns (all ratings = 1) have obvious causes and there is little uncertainty about the chain of events that led to the pathology. Uncertainty about the biological mechanisms underlying the pathology was judged to be high in, for example, rheumatism, rickets, and hiccups (all ratings = 9) whereas burns, broken bones, and thorns are diseases in which mechanisms are more obvious (all ratings = 2-2.5). Mixed effects logistic regression models show that both proxies for uncertainty about the disease predict a higher proportion of religious/supernatural treatments in the diseases ( Fig. 1 ). In the exploratory split of dataset, both uncertainty about the mechanism underlying the illness (OR = 1.76, 89% CI[1.23, 2.52]) and the cause of the illness (OR = 1.63, 89% CI[1.17, 2.33]) predicted more religious/supernatural treatments. In the confirmatory split of the dataset, the effect sizes were similar for uncertainty about the mechanism (OR=1.66, 89% CI [1.10, 2.50]) and cause uncertainty (OR = 1.45, 89% CI[0.96, 2.18]), though the CI crosses 1 for cause uncertainty. Fig. 1. Open in a new tab Supernatural/religious cures for uncertain illnesses. Left panel shows coefficient plot with standardized coefficients from a mixed effect logistic regression. This preregistered analysis supports the hypothesis that diseases that provoke uncertainty would have more religious and supernatural treatments. The center two panels examine the effects of uncertainty on supernatural and religious separately using mixed effect multinomial logistic regression. The Right panel shows the relationship between uncertainty and ritual: this relationship was not predicted. All coefficients are standardized, and models have only one fixed effect. The random effects are disease and school. In the full dataset, which enables the most precise estimate of effect size, uncertainty about the mechanisms underlying the illness (OR = 1.74, 89% CI[1.22, 2.50]) and the cause of the illness (OR = 1.56, 89% CI[1.09, 2.24]) predicted more religious/supernatural treatments. We can interpret this to mean that 1-SD increase in the uncertainty about the mechanism or the cause leads to a 74% and 56% increase respectively in the odds that a supernatural or religious element will be present in the treatment. Note that the full regression tables for these analyses are available in part 2 of SI Appendix . In the logistic regressions above, religious and supernatural treatments are combined and contrasted with naturalistic treatments. In an exploratory mixed effects multinomial logistic regression model, we investigated whether religious and supernatural treatments might have distinct relationships with uncertainty. As the center two panels in Fig. 1 show, we observed a consistent relationship only between uncertainty about the mechanism and supernatural content. The other results directionally suggest that both types of uncertainty may influence both religious and supernatural content, but the findings were not clear. Although we had no theory-driven prediction, diseases with uncertain causes (OR = 1.60, 89% CI [1.23, 2.09] in the full dataset) and mechanisms (OR = 1.82, 89% CI [1.39, 2.38]) were more likely to have ritualistic treatments, see the Right panel in Fig. 1 . Since religious/supernatural cures tend to be ritualistic (as confirmed in our data: OR = 2.77, 89% CI [2.28, 3.37]), we wondered whether ritualistic content could explain the uncertainty-religious/supernatural relationship shown in Fig. 1 . To check this, we reran the mixed effect logistic regression model predicting religious/supernatural vs. naturalistic treatment adding ritualistic (vs. nonritualistic) treatment to the uncertainty predictors with the full dataset. This association between uncertainty about the mechanism and religious/supernatural treatment remained when adding ritualistic vs. nonritualistic treatment to the model (OR = 1.64, 89% CI [1.16, 2.30]), though the association was stronger for the ritualistic predictor (OR = 2.76, 89% CI [2.27, 3.35]) than for the uncertainty about the mechanism. Similarly, the association between uncertainty about the cause and religious/supernatural content remained of a similar magnitude when adding ritualistic treatment to the model (OR = 1.49, 89% CI [1.05, 2.09]). The association was again stronger for the ritualistic predictor (OR = 2.76, 89% CI [2.27, 3.36]). Severity of the Illness. To test H2, which predicted that severe illnesses are more commonly treated with supernatural/religious treatments than with naturalistic treatments, we used five proxies for severity: the proportion of patients expected to die of the disease in 1930s Ireland, the expected level of sickness of the typical patient seeking medical help, the proportion of patients who are very sick, the level of pain of the typical patient seeking medical help, the proportion of patients with severe pain. Diseases rated as severe on most of these metrics included TB, broken bones, cancer, and whooping cough whereas hiccups, ringworm, and warts were rated as more benign. Mixed effects logistic regression models showed no robust relationships between the five proxies of severity and the content of the treatments: for brevity, we report the findings from the full dataset only. One proxy for severity was associated with somewhat higher religious/supernatural content: the sickness of the typical patient (OR = 1.29, 89% CI[0.86, 1.92]). The other four proxies for severity showed no positive or negative effects on supernatural/religious content. These included the proportion of patients expected to die of disease (OR = 1.04, 89% CI[0.79, 1.36]), the proportion of patients with severe pain (OR = 0.82, 89% CI[0.57, 1.19]), the pain of the typical patient (OR = 0.76, 89% CI[0.51, 1.13]) and the proportion of patients who are very sick (OR = 1.09, 89% CI[0.72, 1.63]), see Fig. 2 Left panel. Therefore, we did not find support for the prediction (H2) that severity would be more commonly associated with religious/supernatural treatments than naturalistic treatments. Fig. 2. Open in a new tab Cures for painful and severe illnesses. Left panel suggests that severe diseases are not treated with more supernatural/religious treatments. Top Right panel suggests that painful diseases are less likely to be treated with ritualistic treatments; this was not a predicted association. The Bottom Right panel indicates no relationship between anxiety and ritualistic treatment. These standardized coefficients are derived from mixed effect logistic regression models with one fixed effect and two random effects, disease and school. In an exploratory analysis, we also examined whether more severe diseases had more ritualistic content. Ritualistic treatments were no more common in diseases high in the various measures of severity, with the exception of pain, which predicted lower levels of ritual content. See Fig. 2 , Top Right panel. Anxiety-Inducing Diseases. To test H3, which predicted that ritualistic treatments (both scientific/natural and religious/supernatural) are more common when illnesses provoke high anxiety, we also asked the medical experts to estimate the typical anxiety levels of patients seeking care for this illness as well as the proportion of patients experiencing high anxiety. The results show that the anxiety of the typical patient and the proportion of patients with high anxiety were unreliable predictors of ritualistic treatment in the exploratory, confirmatory, and full datasets. In the full dataset, the OR for the anxiety of the typical patient was 1.02 (89% CI[0.72, 1.43]), and the OR for the proportion of patients with high anxiety was 0.97 (89% CI[0.70, 1.34], Fig. 2 , Bottom Right ). The prediction that (H3) anxiety-inducing diseases would be more commonly associated with ritualistic than nonritualistic treatments, was not supported. A reviewer usefully suggested an alternative way to conceptualize anxiety: it stems from the perception that one’s situation is both highly uncertain and highly consequential. While our measure of anxiety—asking doctors to estimate the anxiety levels of patients with illnesses—presumably captures this emotional state, a more direct approach may be to look at the interaction of uncertainty and severity. As a test case, we regressed uncertainty about the cause and sickness of the typical patient on the probability that a cure had supernatural content. We chose these measures of uncertainty and severity since they had stronger relationships with supernatural content in the aforementioned analyses. The interaction of cause uncertainty and sickness level did not predict more supernatural treatment (OR = 0.80, 89% CI[0.54, 1.18]) and neither did the interaction of other measures of uncertainty and severity. Diseases Elicit Need for Care. We included measures of disability and need for care to test (H4) the material benefits theory described above, which predicted that explicitly religious treatments (as opposed to other types of supernatural content) will be more common than naturalistic treatments when illness results in a need for care. However, neither the disability of the typical patient nor the proportion of patients with severe disability had a robust relationship with religious treatment, with CI including 1 in call cases. In the full dataset the OR estimate for religious treatment vs. naturalistic treatment was 0.26 (89% CI[0.05, 1.33]) for the disability of the typical patient and 0.82 (89% CI[0.62, 1.10]) for the proportion of patients with severe disability. Discussion Cognitive scientists and anthropologists have developed a range of theories about when and why people draw upon religious and supernatural solutions to problems: they bring predictability to uncertain situations, reduce anxiety, and encourage community support and cooperation. By testing this in a corpus of early 20th century Irish folk medicine we found that illnesses with unclear causes or biological mechanisms were more commonly treated with supernatural/religious and ritualistic treatments than disease with more obvious causes or pathologies. This finding is congruent with classical anthropological work from Malinowski and others pointing out the particular importance of supernatural intervention in situations of uncertainty. Jackson et al. ( 41 ) recently found that people across time and societies were more inclined to use supernatural concepts to explain natural events like disease and famine than human-caused events like murder and theft. Illness was the domain most often explained in supernatural terms. Their argument—that the inexplicability of the phenomena is a key determinant of supernatural concepts—is well supported by our more focused exploration of the disease domain. The supernatural cues in our dataset did not appear tightly linked to supernatural entities that had caused the illness. Rather, it seemed that people were seeking practical solutions to problems and when the disease had no intuitive cure, people fell back upon more supernatural concepts. Why should uncertainty about the mechanisms or causes elicit more supernatural content? It seems likely that people want to do something in the event of illness: accepting that a process is governed by unobservable and unmodifiable processes may be psychologically unsatisfying, and in some circumstances, maladaptive. In biomedical contexts, people appear to prefer aversive interventions like surgery over doing nothing (usually termed “watchful waiting”), even when surgical outcomes are poorer ( 42 ). But when the causes and mechanisms of a disease are a mystery, the illnesses present few affordances for action and folk models of biology and healing fail to suggest plausible ways to intervene. Perhaps it is in these contexts that people draw upon more supernatural treatments grounded in more diffuse and elastic conceptual frameworks, such as sympathetic magic. In our dataset, diseases with obvious causes and mechanisms like broken bones tended to get natural treatments (though it should be noted that bone setting was probably one of the few potentially effective treatments in our dataset). The mechanisms/causes of broken and dislocated bones are obvious, and people who hunt or tend animals may well have encountered additional examples of these pathologies. The solutions are intuitive: pull the limb along its axis till the bone recovers its position for a dislocation and stabilize the bone so that it heals straight for a fracture. In the dataset, specific local people with experience of these procedures were mentioned, and unsurprisingly, there were few supernatural cures for broken bones. Cuts, wounds, and burns had similarly natural treatments, though these were unlikely to be effective: cobwebs for cuts and butter/vinegar/paraffin/lard/baking soda for burns. Sties and oral thrush, on the other hand, were thought by the doctors to have causes and mechanisms that would be opaque to the patient, and these tended to have supernatural cures like pointing specific numbers of thorns at the stie or having a person born after the death of their father breath down the thrush-afflicted mouth. If a lack of actionable intervention points is indeed what drives individuals toward supernatural cures, this may clarify why uncertainty about underlying mechanisms is a more reliable predictor of resorting to such remedies than mere ambiguity about what caused the disease. Infectious diseases like mumps, whooping cough, erysipelas (a skin rash), scrofula (a neck swelling, often caused by TB) are more frequently associated with supernatural cures. This pattern could emerge because infectious agents—especially when poorly understood—offer limited points of direct intervention from a lay perspective. Historically, the invisible, dynamic, and often unpredictable nature of infectious agents confounded straightforward causal attributions and thwarted easy, mechanical remedies. Without a clear way to link what we do to preventing or treating these illnesses, ordinary explanations fail to offer practical steps for action. As a result, people may turn to broader, more flexible ideas—like supernatural forces or sympathetic magic—to explain and control diseases they cannot otherwise understand or manage. Our study suggests a more robust relationship between uncertainty and supernatural treatments than between uncertainty and explicitly Christian treatments: It may be that the framing of the questions (“local cures”) may have deterred people from including more orthodox religious solutions since, at the time, there was a tension between orthodox Catholicism and folk Christian/supernatural beliefs. Some, presumably common, religious responses to illness like praying for divine intercession, lighting a votive candle, consulting a priest, or requesting that a Mass be offered were notably absent from the corpus. For similar reasons, the question elicited relatively few orthodox medical solutions (e.g., “seeing a pharmacist”). It is plausible that disease uncertainty and severity could have additional effects on people’s willingness to draw upon orthodox medical or religious solutions that are not well represented in this dataset. Uncertainty about the mechanism and the cause not only predicted supernatural treatments but also ritualistic treatments—or treatments with several repeated steps. A treatment for thrush, for example, was that “the father would put the bill of the gander in the child’s mouth and then squeeze it so that the gander would breathe down the throat of the child. This should be done three mornings in succession.” Or for rickets: “they bathed the child in water taken from a south running stream and then put the child under an ass [donkey] six times.” The most obvious explanation for this is that our operationalization of ritual (having repeated steps) was more a proxy measure of supernaturalness. In contrast, other characteristics of the illness such as their severity, anxiety, and need for support did not have a clear relationship with either supernatural/religious or ritualistic treatments, except their level of pain, which was more commonly treated with nonritualistic treatments. Other previous research has found a relationship between repeated ritualized actions and anxiety reduction ( 24 ). Perhaps this suggests that the relationship does not extend to illness, or that we did not capture anxiety well in our measure. Either way, further research is needed here to draw strong conclusions. We are similarly cautious about overinterpreting the null effect for disease severity, though we have previously found similar null effects for this relationship in another study ( 10 ). It is quite possible that the data recording process—where children solicited cures from older members of their household—introduced biases that masked the effects of severity. Religious cures for TB or cancer might have seemed too heavy a topic for children, for some reason. As we note above, participants may not have considered orthodox religious responses when queried about “local cures.” It seems intuitive that people with a severe illness would be more willing to try a supernatural cure, but this may stem from a broader increase in the willingness to try anything (see ref. 43 ) and this would not necessarily increase the proportion of supernatural cures. This hypothesis is not easily testable in our dataset as many cures repeat or are likely to have a shared origin (for example, there are hundreds of cures for warts that involve snails). The relationship between pain and a reduction in ritualized treatments was not predicted and we have no strong speculations about why this effect exists. We define treatments as ritualistic if they contained steps that were repeated several times over, which may just be difficult for people in pain, or perhaps people in pain would be more impatient with treatments that take time to implement. Conclusion This research can offer some insight into when we reach for supernatural beliefs and thus offer some insight into why we have these beliefs in the first place. We suggest that these beliefs may, in part, arise because of a need to make sense of otherwise confusing or uncertain processes. We focus on illness here, but this tendency may be broader than that and is worth further investigation. Here, we find that uncertainty surrounding an illness makes it more likely that people will reach for supernatural cures. Though we used historical data, these effects are potentially still relevant in how we think about disease today. Alternatives to evidence-based medicine, many of them supernatural in nature, are prevalent everywhere, including contemporary western culture. Uncertainty around an illness may increase the likelihood of people engaging in these types of treatments. Materials and Methods Data. The data for this study was based on the Schools Collection, described previously, hosted by Dúchas.ie ( https://www.duchas.ie/en/tpc ). The Schools Collection corpus was chosen because it offers unique insights into the distribution of supernatural, religious, and naturalistic medical treatments in a religious, preindustrialized population with relatively little access to formal health care. At the time, child mortality was approximately 10% ( 44 ) and life expectancy was 59-y ( 45 ). Ethical approval was not required for this study as the data are in the public domain. We initially included in the study the data from the treatments related to the categorized diseases in the dataset, excluding the uncategorized data tagged with the “folk medicine” label. We included only transcribed cures recorded in English (i.e., excluding Irish language entries) as the coding team did not have the linguistic proficiency in Irish. The categorized diseases were backache, bleeding, boils, broken or dislocated bones, burns, cancer, chilblains, colds, corns, cough, cuts and wounds, earache, erysipelas, fallen breastbone, headache, hiccups, jaundice, measles, mumps, rheumatism, rickets, ringworm, sciatica, scrofula, sore eyes, sore throat, sprains, sties, stings, thorns, thrush, toothache, TB, warts, whitlow, whooping-cough. After web scraping the data, we found out that some of the observations contained multiple cures. We manually separated these cures into different observations. After this, the dataset contained a total of 3,452 cures for the abovementioned diseases. However, some of the diseases contained very few cures (less than 50), which we considered potentially problematic. Consequently, we decided to oversample for these diseases until reaching 50 cures (or the maximum number of cures found on the dataset if this number was below 50) from the uncategorized diseases tagged with the folk medicine label. We eliminated from the dataset fallen breastbone because of its difficulty finding additional cures among the uncategorized treatments and its very low number of cures in the categorized dataset. After doing this, the final dataset contains 3,655 cures for 35 diseases. Seemingly distinct illness categories might, in some instances, reflect different names for the same underlying diseases. While the results below use the preregistered, dataset-derived categories, we reran all the analysis grouping the most similar clusters of illnesses: coughs, colds, and sore-throat were grouped into a respiratory tract infection category, and bleeding and cuts-and-wounds were grouped into a laceration category. The results, shown in part 3 of SI Appendix , are very similar to what we report below. Coding of Treatments. Supernatural and natural. The treatments were classified according to several characteristics. First, the treatments were classified into naturalistic or supernatural treatments. A treatment was considered supernatural if it contained elements that violate commonplace conceptions of causality. The supernatural cures included elements indicating sympathetic magic (i.e., similarity between the cure and the disease, for example, such as using a snail to cure warts); the disease being passed to an object, animal, or person, or acting at distance (e.g., passing warts to another person if they take a bag with stones); accidental events (e.g., finding a snail without looking for it); special categories of people (e.g., the seventh son cures a disease), places (e.g., a halter near the nearest south-running stream), actions (e.g., put a potato in your pocket), incantations (e.g., praying), or other religious elements. A cure was considered naturalistic if it involved none of these supernatural processes and if it included physical contact with the body via ingestion, application, inhalation, or similar. Naturalistic is distinct from evidence-based or effective : rather we coded these cures as naturalistic if there were plausible causal pathways whereby the medical substance could feasibly contact the ailment. This dichotomous variable was the outcome variable to test the hypotheses that predicted that uncertainty and severity are more strongly associated with supernatural than with naturalistic treatments. Religious and not religious. Treatments were classified into religious vs. nonreligious. Religious treatments included elements such as praying, blessing, crossing oneself, or other actions related to Christianity. This variable was later used in tandem with the supernatural vs. naturalistic variable to classify the treatments into three categories: religious, supernatural nonreligious, and naturalistic. We used this variable as the outcome variable to test the prediction that explicitly religious treatments (as opposed to other types of supernatural content) are more common when illness results in a need for care. Ritualization. Treatments were coded for the presence or absence of ritualistic components. In our coding scheme, both naturalistic and supernatural treatments could include ritualistic components. The presence of repetitive elements (e.g., say three Hail Marys) was used as the proxy for presence of ritualistic component as ritualization is often defined as “the limitation of behavioural expressions to predictable stereotypic and repetitive motor patterns” ( 24 ). Initially, we coded the number of repetitions with 0 if there were no repetitions, a number if the specific number of repetitions was specified, and M if the number of repetitions was not specified. Because of the inconsistency with how repetition was recorded (i.e., many cures only specified that the action should be repeated until the cure worked), we transformed this variable into a dichotomous variable of “repetition” or “no repetition.” This transformed variable was used to test the prediction that anxiety is more commonly associated to repetitive/ritualistic treatments than with nonritualistic treatments. Coding process. The classification of the treatments was done by two coders, authors AJ and NR. Initially, both coders coded 210 treatments (5.7%) together to come up with a consistent coding scheme. Then, the coders coded 337 treatments (9.2%) independently to check the consistency of the independent coding (intercoder reliability: Supernatural vs. Naturalistic Cronbach’s a = 0.98; Number of Repetitions Cronbach’s a = 0.96; and Religious vs. Non-religious Cronbach’s a = 0.92). One of the two coders independently classified the remaining treatments in the dataset. Following human coding, we developed large language model (LLM) prompts to categorize the cures again to ensure accuracy. Discrepancies between human and LLM codes were resolved by author MdB. Note that the LLM coding was not preregistered as LMMs were not sufficiently advanced when we were developing the protocol. There were no substantive changes to the results as a result of adding this step. Additional details about the coding, including the LLMs and prompts used, are available in part 4 of SI Appendix . Coding of diseases. Two medical doctors graded each of the diseases in the dataset. One was a general practitioner/family doctor with 18 y of postqualification experience practicing medicine, largely in rural Irish surgeries. The other was a pediatric consultant with a similar duration of experience. The two doctors were first briefed on the broad goals of the project. The hypotheses were not described. They were then asked to answer the questions “assuming no treatment and public health standards of 1900–1930s Ireland” and were provided with information about infant mortality rates, public health infrastructure, and medical standards at that time. The doctors were free to consult any resources they might find helpful. In answering questions about the illnesses, they were asked to imagine that the distribution of patients included quite minor as well as severe cases. The assessments of each disease by the doctors included 11 questions ( Table 1 ). Table 1. Correlation between the ratings of different characteristics of the disease given by two medical doctors Question asked about each of 35 diseases Interdoctor correlations Prediscussion Postdiscussion Proportion of patients who die of disease assuming no treatment. 0.63*** 0.98*** Sickness of typical patient seeking for help (9-point Likert scale from 1 [no sickness] to 9 [close to death]). 0.50** 0.85*** Proportion of patients who are very sick (i.e., a rating of 7 or above on the scale for the previous question). 0.72*** 0.80*** Typical Disability of patient seeking for help (9-point Likert scale from 1 [no reduction in capacities or assistance needed] to 9 [unable to perform basic functions, totally dependent on others]). 0.55*** 0.81*** Proportion of patients with severe disability (i.e., 7 or above on the previous questions) for 1 mo or more. 0.60*** 0.75*** Pain of typical patient seeking for help (9-point Likert scale from 1 [no pain at all] to 9 [the worst pain possible]). 0.52** 0.86*** Proportion of patients with severe pain (i.e., 7 or above in the previous question on the day they seek help). 0.42* 0.68*** Uncertainty about the cause of the disease (9-point Likert scale from 1 [Likely to be very certain about the antecedent cause of the disease] to 9 [Likely to be very uncertain about the antecedent cause of the disease]). 0.65*** 0.93*** Uncertainty about the mechanism of the disease (9-point Likert scale from 1 [Likely to be very certain about the biological problem causing their symptoms] to 9 [Likely to be very uncertain about the biological problem causing their symptoms]). 0.11 0.96*** Anxiety of typical patient seeking for help (9-point Likert scale from 1 [not anxious/distressed at all] to 9 [extremely anxious/distressed]). 0.88*** Proportion of patients with high anxiety (i.e., level 7 or above on the previous anxiety scale) 0.89*** Open in a new tab * P < 0.05, ** P < 0.01, *** P < 0.001. These questions were first answered by the doctors independently. The correlations coefficients are provided in Table 1 . The doctors then met in person to examine each other’s codes, explain their reasoning to one another and, if they wished, adjust their ratings. They were not encouraged to reach consensus, but to change rating if they found each other’s arguments convincing. Discrepancies in the initial coding for some items stemmed from factors like unfamiliarity with the disease among one of the doctors or differing professional experience (as a consultant pediatrician, one doctor tended to see more severe cases). Following discussion, agreement was substantially higher; see Table 1 . We then averaged the estimates from the two doctors. Note that two questions about anxiety levels were introduced after the doctor’s initial independent coding and was coded just once by each doctor during the consensus meeting. The pairwise correlations between the different disease questions are shown in part 5 of SI Appendix . Data Analysis. Statistical analyses were preregistered in two parts. In the first part of the preregistration ( https://osf.io/ed8v9 ), we outlined the exploratory analyses, produced with 50% of the dataset, on the relationship between disease characteristics and cure characteristics. The models derived from these exploratory analyses (including the R script) were preregistered in the second part of the preregistration ( https://osf.io/rup2d ) to be used as confirmatory analyses of the results with the second 50% of the dataset. As stated in the second part of the preregistration, we have also tested the hypotheses with the same models produced with 100% of the dataset to attain more reliable estimates of the effects. All the statistical analyses were conducted with R 4.1.0. To control for the effects of the shared variation within diseases and schools, we used mixed effects regression modeling with disease and school as separate random effects. The outcome variables for testing if causal opacity or severity produced more supernatural treatments (hypothesis 1 and 2; supernatural vs. naturalistic treatment) and ritualization through repetition reduced anxiety (hypothesis 3; with repetition vs. no repetition) were dichotomous. Therefore, we ran mixed effects logistic regression models with the lme4 package ( 46 ) to test these hypotheses. The outcome to test if religious treatments were more common in diseases needing care (hypothesis 4) were three categories (religious, supernatural nonreligious, or naturalistic treatment). Therefore, we ran mixed effects multinomial regression models with the mclogit package ( 47 ) to test this hypothesis. Because we used proxies of the variables of interest in two different scales (i.e., proportion and Likert scale from 1 to 9), all the coefficients reported here are standardized. We used 89% CI (see ref. 48 ) throughout. Supplementary Material Appendix 01 (PDF) pnas.2511006122.sapp.pdf (872.1KB, pdf) Acknowledgments We thank Dr. Aoife Carrol and Dr. Tadgh de Barra for rating the diseases, Ms. Arafa Said for assistance with the AI prompts, and Dr. Hugh Turpin for discussions on the historic religious context in Ireland. This work was supported by The Issachar Fund and the Templeton Religion Trust (grant number TRT0207). Author contributions M.d.B. and A.K.W. designed research; A.V.J. and N.R. performed research; M.d.B., A.V.J., and A.K.W. analyzed data; and M.d.B., A.V.J., and A.K.W. wrote the paper. Competing interests The authors declare no competing interest. Footnotes This article is a PNAS Direct Submission. Data, Materials, and Software Availability Cures and their coding plus the analysis syntax data have been deposited in figshare ( https://figshare.com/s/77c9c858581e5014feed ) ( 49 ). Supporting Information References 1. Murdock G. P., Wilson S. F., Frederick V., World distribution of theories of illness. Ethnology 17, 449–470 (1978). [ Google Scholar ] 2. Watson-Jones R. E., Busch J. T. A., Legare C. H., Interdisciplinary and cross-cultural perspectives on explanatory coexistence. Top Cogn. Sci. 7, 611–623 (2015). [ DOI ] [ PubMed ] [ Google Scholar ] 3. Obeid T., Abulaban A., Al-Ghatani F., Al-Malki A. R., Al-Ghamdi A., Possession by ‘Jinn’as a cause of epilepsy (Saraa): A study from Saudi Arabia. Seizure 21, 245–249 (2012). [ DOI ] [ PubMed ] [ Google Scholar ] 4. Labotka D., Gelman S. A., Scientific and folk theories of viral transmission: A comparison of COVID-19 and the common cold. Front. Psychol. 13, 929120 (2022). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Legare C. H., Gelman S. A., Bewitchment, biology, or both: The co-existence of natural and supernatural explanatory frameworks across development. Cogn. Sci. 32, 607–642 (2008). [ DOI ] [ PubMed ] [ Google Scholar ] 6. Raman L., Gelman S., A cross-cultural developmental analysis of children’s and adults’ understanding of illness in South Asia (India) and the United States. J. Cogn. Cult. 4, 293–317 (2004). [ Google Scholar ] 7. Raman L., Winer G. A., Children’s and adults’ understanding of illness: Evidence in support of a coexistence model. Genet. Soc. Gen. Psychol. Monogr. 128, 325–355 (2002). [ PubMed ] [ Google Scholar ] 8. Malinowski B., Magic, Science and Religion and Other Essays (Beacon Press, 1948). [ Google Scholar ] 9. Singh M., The cultural evolution of shamanism. Behav. Brain Sci. 41, e66 (2018). [ DOI ] [ PubMed ] [ Google Scholar ] 10. Willard A. K., Rosun N., de Barra M., God, witchcraft, and beliefs about illness in Mauritius. Religion Brain Behav., 10.1080/2153599X.2024.2363748 (2024). [ DOI ] [ Google Scholar ] 11. Hong Z., Chance as a (non)explanation: A cross-cultural examination of folk understanding of chance and coincidence. Psychol. Rev., 10.1037/rev0000568 (2024). [ DOI ] [ PubMed ] [ Google Scholar ] 12. Kay A. C., Gaucher D., McGregor I., Nash K., Religious belief as compensatory control. Pers. Soc. Psychol. Rev. 14, 37–48 (2010). [ DOI ] [ PubMed ] [ Google Scholar ] 13. Hong Z., Henrich J., The cultural evolution of epistemic practices: The case of divination. Hum. Nat. 32, 622–651 (2021). [ DOI ] [ PubMed ] [ Google Scholar ] 14. Evans-Pritchard E. E., Witchcraft, Oracles and Magic among the Azande (Clarendon Press, Oxford, 1937). [ Google Scholar ] 15. Hagger M. S., Koch S., Chatzisarantis N. L. D., Orbell S., The common sense model of self-regulation: Meta-analysis and test of a process model. Psychol. Bull. 143, 1117 (2017). [ DOI ] [ PubMed ] [ Google Scholar ] 16. Jong J., Halberstadt J., Bluemke M., Foxhole atheism, revisited: The effects of mortality salience on explicit and implicit religious belief. J. Exp. Soc. Psychol. 48, 983–989 (2012). [ Google Scholar ] 17. Becker E., The Denial of Death (Macmillan Publishing, 1973). [ Google Scholar ] 18. Ellis L., Wahab E. A., Religiosity and fear of death: A theory-oriented review of the empirical literature. Rev. Relig. Res. 55, 149–189 (2013). [ Google Scholar ] 19. Sibley C. G., Bulbulia J., Faith after an earthquake: A longitudinal study of religion and perceived health before and after the 2011 Christchurch New Zealand earthquake. PLoS One 7, e49648 (2012). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Norenzayan A., Hansen I. G., Belief in supernatural agents in the face of death. Pers. Soc. Psychol. Bull. 32, 174–187 (2006). [ DOI ] [ PubMed ] [ Google Scholar ] 21. Reed P. G., Religiousness among terminally ill and healthy adults. Res. Nurs. Health 9, 35–41 (1986). [ DOI ] [ PubMed ] [ Google Scholar ] 22. Idler E. L., Kasl S. V., Hays J. C., Patterns of religious practice and belief in the last year of life. J. Gerontol. B Psychol. Sci. Soc. Sci. 56, S326–S334 (2001). [ DOI ] [ PubMed ] [ Google Scholar ] 23. Meacham F., Bergstrom C. T., Adaptive behavior can produce maladaptive anxiety due to individual differences in experience. Evol. Med. Public Health 2016, 270–285 (2016). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Lang M., Krátký J., Shaver J. H., Jerotijević D., Xygalatas D., Effects of anxiety on spontaneous ritualized behavior. Curr. Biol. 25, 1892–1897 (2015). [ DOI ] [ PubMed ] [ Google Scholar ] 25. Brooks A. W., et al. , Don’t stop believing: Rituals improve performance by decreasing anxiety. Organ. Behav. Hum. Decis. Process. 137, 71–85 (2016). [ Google Scholar ] 26. Whitehouse H., The Ritual Animal: Imitation and Cohesion in the Evolution of Social Complexity (Oxford University Press, 2021). [ Google Scholar ] 27. Eilam D., Zor R., Szechtman H., Hermesh H., Rituals, stereotypy and compulsive behavior in animals and humans. Neurosci. Biobehav. Rev. 30, 456–471 (2006). [ DOI ] [ PubMed ] [ Google Scholar ] 28. Norenzayan A., et al. , The cultural evolution of prosocial religions. Behav. Brain Sci. 39, e1 (2016). [ DOI ] [ PubMed ] [ Google Scholar ] 29. Johnson D., God Is Watching You: How the Fear of God Makes Us Human (Oxford University Press, 2015). [ Google Scholar ] 30. Storm I., Does security increase secularity? Evidence from the British Household Panel Survey on the relationship between income and religious service attendance. J. Relig. Eur. 10, 328–349 (2017). [ Google Scholar ] 31. Kavanagh C., Jong J., Whitehouse H., “Ritual and religion as social technologies of cooperation” in Culture, Mind, and Brain: Emerging Concepts, Models, and Applications, Kirmayer L. J., Worthman C. M., Kitayama S., Lemelson R., Cummings C. A., Eds. (Cambridge University Press, 2020), pp. 325–361. [ Google Scholar ] 32. Stark R., The Rise of Christianity: A Sociologist Reconsiders History (Princeton University Press, 1996). [ Google Scholar ] 33. Sugiyama L. S., Illness, injury, and disability among Shiwiar forager-horticulturalists: Implications of health-risk buffering for the evolution of human life history. Am. J. Phys. Anthropol. 123, 371–389 (2004). [ DOI ] [ PubMed ] [ Google Scholar ] 34. Sosis R., Does religion promote trust?: The role of signaling, reputation, and punishment. Interdiscip. J. Res. Relig. 1, 2–30 (2005). [ Google Scholar ] 35. Barra M., Cownden D., Jansson F., Aversive medical treatments signal a need for support: A mathematical model. Evol. Hum. Sci. 1, e4 (2019). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Bliege Bird R., Ready E., Power E. A., The social significance of subtle signals. Nat. Hum. Behav. 2, 452–457 (2018). [ DOI ] [ PubMed ] [ Google Scholar ] 37. The Schools’ Collection, Dúchas © National Folklore Collection, UCD is licensed under CC BY-NC 4.0. (2013). https://www.duchas.ie/en/info/contact . Accessed 20 April 2021. 38. Ú. Bhreathnach, “An introduction to the National Folklore Collection and Meitheal Dúchas.ie” in Irish Islands, Folk Studies and Folklore: Research Papers in Archaeology, Folklore and History, Higgins J., Ed. (The Crows Rock Press, Galway, 2019), pp. 97–106. [ Google Scholar ] 39. Wootton D., Bad Medicine: Doctors Doing Harm Since Hippocrates (OUP Oxford, 2007). [ Google Scholar ] 40. de Barra M., “Cultural evolution of ineffective medicine” in The Oxford Handbook of Cultural Evolution, Tehrani J. K., Kendal J., Kendal R., Eds. (Oxford University Press, 2023). [ Google Scholar ] 41. Jackson J. C., et al. , Supernatural explanations across 114 societies are more common for natural than social phenomena. Nat. Hum. Behav. 7, 707–717 (2023). [ DOI ] [ PubMed ] [ Google Scholar ] 42. Fagerlin A., Wang C., Ubel P. A., Reducing the influence of anecdotal reasoning on people’s health care decisions: Is a picture worth a thousand statistics? Med. Decis. Making 25, 398–405 (2005). [ DOI ] [ PubMed ] [ Google Scholar ] 43. Foster K. R., Kokko H., The evolution of superstitious and superstition-like behaviour. Proc. R. Soc. B: Biol. Sci. 276, 31–37 (2009). [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. McGovern M. E., Publication: Progress and the lack of progress in addressing infant health and infant health inequalities in Ireland during the 20th century. J. Stat. Soc. Inquiry Soc. Ireland, 48, 117–145 (2016). [ Google Scholar ] 45. Walsh B., Life expectancy in Ireland since the 1870s. Econ. Soc. Rev. (Irel) 48, 127–143 (2017). [ Google Scholar ] 46. Bates D., Mächler M., Bolker B., Walker S., Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48 (2015). [ Google Scholar ] 47. Elff M., mclogit: Multinomial logit models, with or without random effects or overdispersion. Biometrics (2022). 48. McElreath R., Statistical Rethinking: A Bayesian Course with Examples in R and Stan (Chapman and Hall/CRC, 2018). 49. de Barra M., Dataset and analysis syntax: Mysterious illnesses have supernatural and ritualistic cures: Evidence from 3,655 century-old Irish folk cures. Figshare. 10.6084/m9.figshare.28794950.v2. Deposited 15 April 2025. [ DOI ] [ PMC free article ] [ PubMed ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Appendix 01 (PDF) pnas.2511006122.sapp.pdf (872.1KB, pdf) Data Availability Statement Cures and their coding plus the analysis syntax data have been deposited in figshare ( https://figshare.com/s/77c9c858581e5014feed ) ( 49 ). Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences ACTIONS View on publisher site PDF (273.3 KB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Irish Health Survey (IHS), 2025#291999
• Health Survey Northern Ireland, 2013-2014#331796
• Introduction to Laleh Bakhtiar's,_The Sufi Enneagram: The Secrets of the Symbol Unveiled_#86973
• Anthropological reflections on missiological issues#372470
• Experiences of dementia in farming communities: A scoping review#179821