1. Introduction
2. What Does It Mean for Prayer to Be Effective?
3. Hypotheses about the Efficacy of Prayer
4. Methodological Problems
5. Retrospective Observational Studies
6. Experimental Studies
7. Prospective Observational Studies
8. Discussion
Secular discussions of the efficacy of prayer suffer from deficient understanding of orthodox Christian theology and the science of metaphysics. Lack of theological understanding results in defining hypotheses about prayer that are altogether alien to Christian doctrine, so that we are testing a straw man’s theory of prayer. Lack of metaphysical understanding results in untenable a priori conjectures about how prayer can or cannot work, or how God must or must not behave. As most modern scientists deny that there even is a science of metaphysics, it is especially strange that some should declare anything to be metaphysically impossible.
Aside from the problems of defining our hypotheses and axioms, there remain serious methodological problems with any study intending to test hypotheses about the efficacy of prayer. It may seem obvious to scientists that randomized controlled interventional studies are preferable to observational studies, to avoid selection bias, but in the case of prayer studies it is practically impossible to have adequate experimental controls. One cannot prohibit control subjects from praying for themselves, as this is both impractical and unethical. Further, the very notion of a prescribed prayer for the express purpose of testing efficacy is utterly contrary to the piety and humility that are the preconditions of Christian prayer. Lastly, there is a problem of quantification in the materialist paradigm of a dose-response model. The notion that prayer should be more effective depending on its duration or on the number of people praying is theologically dubious, to say the least. Yet if we cannot prevent control subjects from praying for themselves, all studies of intercessory prayer are implicitly adopting a dose-response paradigm, testing if more is better.
The methodological problems cut both ways. There could be real effects of prayer that are undetectable due to the study design, or there could be false correlations showing an effect that is actually due to selection bias. Some propose controlling
for the physically salutary behaviors of the devout (abstinence from narcotics, smoking, etc.), but that depends on what we are testing. If we wish to capture all the benefits of prayer, including the spiritual benefits and its behavioral effects, it would be misguided to control
for these behavioral differences. If, on the other hand, we are trying to find some causation that is indepndent of the person’s interior disposition, controlling for such differences makes sense, but then it is unclear if we are really testing the efficacy of prayer in the sense of Christian doctrine.
The strong claims that Christians have made about the efficacy of prayer seem to suggest that this should be empirically observable and measurable. Even if prayer for temporal goods is effective
only some of the time, that disparity in outcomes ought to appear in at least some observational studies, regardless of how the outcome is achieved, i.e., providentially or miraculously.
We will examine orthodox Christian doctrine on the efficacy of prayer, in order to define hypotheses that accurately reflect Christian claims. We will then see to what extent past studies on prayer, whether observational or interventional, are able to test these hypotheses, keeping a close eye on their methodological limitations.
We should first clarify what is meant by the efficacy of prayer in orthodox Christian doctrine. This breaks down into several questions.
Prayer in the most general sense is directing our attention to divine things, though we are concerned in particular with prayers of petition. All prayer, even that of thanksgiving where we do not ask for anything, has the effect of promoting fellowship or communion with God, and aiding us in our salvation. Indeed, the principal object and effect of prayer is to lead us toward salvation, and all other effects are subordinate to that end. Thus the Christian should pray primarily for spiritual goods, though it is also licit and good to pray for temporal goods. In the latter case, however, such goods may be granted or denied, depending on whether they are expedient to our salvation. We may pray for ourselves or for others, but in the latter case, since others cannot deserve our merits, there is no guarantee that they will receive even spiritual benefits. Thus St. Augustine lays down these four conditions for absolutely efficacious prayer:
(1) We pray for ourselves
(2) things not contrary to our salvation
(3) piously
(4) and perseveringly.[1]
God may heed other prayers as well, but it is only when these four conditions are met that we are guaranteed always to obtain what we request, though perhaps not immediately. This is what is meant by Christ’s absolute promise to the apostles and all the saints: Amen, amen I say to you: if you ask the Father any thing in my name, He will give it to you.
(John 16:23)
We should not expect this absolute efficacy to be testable empirically, since we cannot measure spiritual goods, and we cannot know if the temporal goods we request would actually be good for our salvation or not. Moreover, the condition of piety is incompatible with requests designed to put God to the test, or to seek material benefits without regard for the spiritual. The pious are more likely to pray for the health and temporal goods of others rather than their own. Lastly, the condition of perseverance means we cannot place short temporal limits on our observations.
Christ recognized that most people need external signs of divine power in order to be assured of promises of invisible blessings. Thus he worked many miracles healing the body, as an external sign of his power to cure the soul. As such signs are no longer necessary to proclaim Christ’s mission, we do not expect petitions for such miraculous healings to be granted always or even most of the time. Nonetheless, even in modern times, there are many well documented cases of apparently miraculous healings in response to prayer. Whether these healings are properly supernatural or occur by providential disposition of natural forces, they are believed in some sense to be the result or effect of prayer. In principle, such a hypothesis ought to be empirically testable.
Does the act of prayer itself cause these effects, or at least contribute to them in some way? At first glance, this seems impossible, insofar as prayer is a human act, and we request things that are beyond human power to effect. Clearly, divine power is required, for indeed prayer directs us to recognize dependence on God for all spiritual and temporal goods. Nonetheless, if we believe that God is truly answering
prayer, then the prayer itself belongs somewhere among contributing factors that cause the effect. This seems to imply that the effect would not have occurred had the prayer not been made, which in turn seems to imply that God has acted in consequence to a human decision, which appears to be incompatible with the immutability of the divine Will.
St. Thomas solves the difficulty as follows:
…Divine providence disposes not only what effects shall take place, but also from what causes and in what order these effects shall proceed. Now among other causes human acts are the causes of certain effects. Wherefore it must be that men do certain actions, not that thereby they may change the Divine disposition, but that by those actions they may achieve certain effects according to the order of the Divine disposition: and the same is to be said of natural causes. And so is it with regard to prayer. For we pray not that we may change the Divine disposition, but that we may impetrate that which God has disposed to be fulfilled by our prayers in other wordsthat by asking, men may deserve to receive what Almighty God from eternity has disposed to give,as Gregory says (Dial. i, 8).[2]
God is inalterably disposed from eternity to grant a person a certain benefit if he should only pray for it. This conditioned will does not make God alterable or dependent on a creature, for the condition is imposed by God, not by the creature, and God foreknows eternally what the person will choose. The man who prays thereby meets God’s freely determined condition for granting the benefit, and in that sense his prayer truly is a cause of the benefit. This human cooperation with Divine Providence avoids the fatalistic error that prayer is useless, either because everything happens of necessity, independently of Providence, or else that Divine Providence, because it is unchangeable, has no regard for human choices.
Christians hold that everything is governed by Divine Providence, which may be broadly described as an act of practical wisdom whereby God orders all events in the universe, if not always by coercion, at least by His indispensable concurrence. Thus absolutely everything, even ordinary natural processes following intelligible laws or properties of created beings, is ultimately governed by Divine Providence.
In what sense, if any, is the Providential granting of prayers distinct from the Providential governance of the natural order? Often there is no necessity that God should alter or suspend the natural order to grant a prayer. He needs only to ordain from eternity that specific physical conditions in the world will exist, naturally resulting in the desired outcome. Since God foreknows from eternity whether or not a person will in fact pray for this benefit, He may dispose the world appropriately, long before the prayer occurs in time. It is unclear whether such disposition should result in any statistical difference from naturally expected distributions.
God may also, to be sure, suspend or override the natural order by performing a properly supernatural miracle, in honor of a pious prayer. The good and rational government of the universe requires that this be done sparingly, lest we should become as incapable as children, constantly excused from the natural consequences of our actions and the intelligible actions of other creatures. At any rate, it is certainly the case that miracles are rare, even if it could be proved that every reputed miracle were genuine.
A last consideration is whether the answering of prayers is something granted only to Christians or some other clearly identifiable class of persons. St. Thomas professes that…
…God hears a sinner’s prayer if it proceed from a good natural desire, not out of justice, because the sinner does not merit to be heard, but out of pure mercy, provided however he fulfil the four conditions given above, namely that he beseech for himself things necessary for salvation, piously and perseveringly.[3]
The prayer of someone who lacks the state of grace can never have strict or even condign merit, so he cannot deserve in either sense the gifts he requests by prayer. Nonetheless, God may grant his impetration or entreaty out of mercy, provided it is for the good of his salvation, etc. Since there is no condign merit in the sinner’s prayer, it need not be always heard, even when the four conditions are met, for God is not obligated by His Justice to grant it.
Since there are no limits on the scope of Divine Mercy, which is a free gift, there are no persons among the living whose prayers might not be granted. So we cannot measure the efficacy of prayer by assuming that only Christians, or only those in the state of grace, may have their prayers heard. There is a disparity in that only Christians in the state of grace are capable of the kind of prayer that is absolutely efficacious, but it is problematic for an objective observer to determine when that kind of prayer is being performed.
Given the Catholic doctrinal tradition about whether and in what sense prayer can be efficacious, we may attempt to formulate some hypotheses about how the efficacy of prayer may manifest itself in observable reality. Not all of these hypotheses may be empirically testable, for methodological reasons we will discuss later.
It must be admitted that Christian and specifically Catholic claims about prayer often greatly exceed the properly doctrinal claims. Many a cleric or layperson will extol the efficacy of prayer in superlatives, advising the faithful that prayer is the most effective means for attaining just about everything. Charities will say that receiving prayers is more important than monetary contributions. Entire parishes are urged to pray for various causes, from the health of a parishioner to the good of the nation, as though there were strength in numbers of prayers. When someone is ill, we are counseled that prayer is the strongest medicine. Moreover, various traditional prayers, especially Marian prayers such as the Rosary and the Memorare, are said to have absolute efficacy if they are prayed piously or under certain conditions, or sometimes without mentioning any conditions. These popular claims, if taken literally, would apply Christ’s promise of absolute efficacy to prayers well beyond the scope defined by the Augustinian-Thomist doctrinal tradition.
Anyone who has lived long will surely encounter contradictions of this pervasive popular tradition, as even the most pious prayers may go unanswered. Often, the explanations for such failures compound error with error, as we are told, like Job, that we must be guilty of some hidden impiety, or that the evil that God permits must actually be deserved or for our good. While some may actually find consolation in these rationales, for others it can be a needless source of spiritual agony, and for those with strong critical faculties, it may sound like special pleading that only brings discredit on claims about the efficacy of prayer.
Our hypotheses about the efficacy of prayer should encompass these popular traditions as well, though we should distinguish when we are dealing with doctrinal versus popular claims.
The hypotheses grounded in doctrine are mostly untestable empirically, if not in principle then in practice. Hypotheses A1 and A2 are untestable in principle, since we cannot determine by externals whether some temporal good is conducive to salvation, and we cannot measure spiritual goods. A3, A4, and A5 may be testable in principle, but the practicality of testing them may be limited by methodological constraints, while A6 and A7 may negate our ability to detect effectiveness by statistical means.
The popular hypotheses B1 and B2 are much more likely to be testable. B3 is testable only insofar as the claim of effectiveness is not subject to immeasurable conditions. Even if our claims about the efficacy of prayer should be more modest than those made by popular piety, we may expect to find at least some measurable effect, regardless of the means by which that effect is realized, as long as we can overcome methodological difficulties.
Modern scientific methodology is designed to detect regular correlations among sensible phenomena, using observational studies or controlled experiments. In biomedical studies, the effectiveness of an intervention is measured by statistical analysis of outcomes in many subjects. The application of such methodology to studies of prayer can be deeply problematic, as this methodology can entail assumptions that are incompatible with the mentioned hypotheses about prayer. Even when it is possible in principle to test a hypothesis about prayer scientifically, there may be practical or ethical problems that impede adequate study design.
The most glaring difficulty is that scientific methodology operates on a mechanistic paradigm, dealing as it does with primarily inanimate or at least unintelligent natures, which can be analyzed in terms of purely quantitative laws. Science works because most natural objects behave in predictable ways according to intelligible extrinsic laws. If we lived in a magical world where every being could act as it pleased or be subjected to the will of some remote person, the behavior of natural objects would defy analysis, and we could never be sure that the immediate cause was anywhere in the vicinity of the effect. Yet Christian belief about prayer involves the exercise of free will, both by the person who prays, and God who chooses whether or not to grant what is entreated. A free and sovereign will cannot be subjected to any law or necessity.
Christians and Jews have long recognized that it is undignified, even blasphemous, to attempt to subject the Divine Will to any kind of test, as if God could be compelled by man. Thus the condition Thy Will be done
is always at least implicit. Any sort of test or intervention that would presume to force God to choose A or B cannot expect divine cooperation. This is clearly a different scenario from ordinary science, where controlled experiments can indeed force information out of a system.
The Divine Will is knowable to humans only in a generic way, so we know that God wills what is good and just. Yet we cannot usually know in advance what is the Divine Will on some determinate matter. We cannot know in advance what particular goods God wills for us, though we are enjoined to pray for spiritual goods and those temporal goods (daily bread
) that are material necessities or conducive to salvation.
The pagans sought to discern a divine will through divination, usually by examining some natural process involving chance, e.g., reading tea leaves or animal entrails. These practices were condemned by Jews and Christians as they ascribe divine power to natural objects. Nonetheless, there are some famous instances of apparently sanctioned divination in the Bible, such as the use of Urim and Thummim by the high priests of Israel, or the drawing of lots to replace Judas among the Apostles. Here, however, the intent was not to compel God to prove Himself by acting a certain way (as at Massah, referenced by Deut. 6:16) or to obtain knowledge of the future, but to beseech God to decide what ought to be done on a difficult spiritual matter. Even this much was not dared except after praying for guidance. The most pious prayer cannot compel God to manifest His will on some determinate matter, and any attempt to do so would be illicit divination.
The notion of effectiveness ordinarily used in science is different from the kind of efficacy ascribed to prayer. When we say aspirin is effective for relieving headaches, we mean there is a physically causal mechanism whereby aspirin accomplishes this effect, and that the effect in a given subject would not have occurred were it not for the aspirin. The intervention of the aspirin changes the outcome from what it would have been otherwise, as is proven by comparison with control subjects.
Prayer, by contrast, has efficacy in an entirely different sense. God knows from eternity what we will pray, and in honor of this prayer, He freely and eternally wills the corresponding effect. Since the effect is realized by inscrutable and inalterable Providence, it is far from obvious that prayer changes the outcome in some empirically measurable sense that can be compared against controls. Divine Providence has disposed of all outcomes from eternity, foreknowing all our choices, and it is hardly necessary that such disposition should require deviation from naturally expected probability distributions. To be sure, God can dispose that prayer is truly the cause of the effect, but this is not by some mechanistic or even temporal connection.
Studies on the efficacy of prayer tend to focus on prayers for temporal goods, such as a health outcome, since these results can be measured objectively. Yet prayer for temporal goods has the lowest priority in Christianity, and even then it is subordinate to its expediency in realizing spiritual goods.
Effectiveness in biomedical studies is measured by statistical differences in outcomes with or without an intervention. Yet if all we study are prayers for temporal goods, without regard for expediency toward the spiritual good, it is not clear that we should expect any such measurable difference. Christianity is not a materialistic religion that brings health and prosperity independently of spiritual considerations. Studies expecting such an outcome come from a secular perspective that sees no difference between Christian and pagan prayer.
Nonetheless, we might expect there to be at least a small statistical difference in temporal outcomes in large observational studies (which do not put God to the test), as it may be expected that a small subset of subjects will pray piously and perseveringly for a temporal good that benefits their salvation. Even this, however, requires a supposition that Divine Providence will act in such a way as to create a distribution of outcomes different from what would be expected by nature.
This last point is important, because many studies seem to assume that prayer must be efficacious in a manner analogous to natural causes, so that it is an extra agent on top of natural agents. This is properly the case only when a prayer is realized by a supernatural miracle. Yet miracles are so infrequent as to be statistically insignificant. The credibility of particular miracles can only be ascertained forensically, not scientifically, as these cannot be repeated on demand.
More generally, there is a repeatibility problem in prayer studies. While we can certainly design studies that can be repeated by other investigators, there can be no expectation of similar results in similar trials. There is no reason why the frequency with which God grants or does not grant prayer should have any constancy over time or place. Indeed, most religious traditions recount times when their gods had temporarily forsaken them, ignoring their prayers and sacrifices. While Christians have a more optimistic view of the true God, even they acknowledge that He may sometimes ignore their prayers, either in correction of some offense or for some inscrutable reason. Indeed, we find a broad inconsistency in results among prayer studies, and they might not all be attributable to methodological differences.
The materialistic biomedical paradigm causes some investigators to adopt a dose-response model of prayer, as if more prayers should be more effective than fewer prayers. This model is implicitly adopted by all intercessory prayer studies. Since it is practically certain that most people in both the intervention and control groups will pray for themselves or have family praying for them, the intercessory prayer is an add-on
that will supposedly enhance effectiveness. This again mistakes prayer for magic. We have noted, on the contrary, that Christian doctrine regards prayer for others as less effective than prayer for oneself, for only in the latter case can prayer have condign merit.
Arguably, there has never been a prayer study with adequate controls, at least not by standards of ordinary biomedical studies. Investigators cannot ethically prohibit control subjects from praying for themselves or having families pray for them. Ignoring this is justifiable only if we have a dose-reponse model of intercessory prayer. Otherwise, we have a completely invalid study that would not be published by a scientific journal in any other context. If we are testing the effectiveness of a drug, but allow the placebo group to take unspecified amounts of the drug, the study would be a joke. Any reported effect or non-effect would be meaningless.
Another problem is that we cannot tell who is really getting the intervention, as it is impossible to distinguish by external measures between effective and ineffective practitioners of prayer. Outward holiness does not suffice, for one’s own holiness cannot confer condign merit by intercessory prayer. On the other hand, God may grant the prayer of even the most overt sinner. Some studies try to get around this problem by expanding their scope to several prayer groups of diverse religions, in the hope that at least some are effective and will have a measurable result.
While we cannot be optimistic about seeing measurable effects in objective temporal outcomes, especially in interventional studies that put God to the test,
there is a better expectation of finding positive effects of the subjective experience of prayer. In intercessory prayer studies, this means telling some subjects that they are being prayed for, and seeing if that fact alone improves psychological or physiological health. Prayer may also confer psychological or physiological health benefits, from the mere act of contemplation, or perhaps from salutary lifestyle choices such as abstinence from drugs or strong drink. Sometimes we may control for the latter if we wish to focus on the effectiveness of prayer as such, rather than on the benefits of religiosity in general.
The most important benefits of prayer, the spiritual benefits, are undetectable to science. Studies of temporal benefits are sometimes motivated by a desire to find an external sign of the supernatural efficacy of prayer. It is unclear, however, if the methodological difficulties outlined can ever be overcome. The doctrinally defined hypotheses are most difficult to test empirically. The popular hypotheses, while more testable, rely on dubious claims about prayer that are disconnected from Christian theology. As we examine some significant prayer studies, we will find that most are better suited to testing the latter hypotheses than the former, though all have various methodological problems.
Remarkably, most of the basic tools of statistical analysis taken for granted in scientific research were not invented until the late nineteenth and early twentieth centuries, though modern empirical science had by then existed for centuries. Early modern science had focused on the law-like behavior of inanimate substances in physics and chemistry, but an expansion of interests into biology, sociology and psychology required analytic tools for processes that did not invariably produce the same effect. Before the development of modern statistics, it could not have occurred to anyone that the efficacy of prayer might be testable by this method. There had been disputes about the scientific possibility of miracles, but these rested on the particulars of certain events, or on claims that physical law admitted no exceptions, rather than statistical analyses of efficacy.
By the Victorian era, British thought had largely adopted some version of the Reformed distinction between general and special providence. By general providence, God disposed of all creation in general via the natural order, creating natural beings and their properties and laws. By special providence, God was solicitous of the good of the Church and its people, and sought to reward or punish them by special interventions, though not necessarily violating any law of nature. Calvin had cautioned that one could not infer the Divine disposition from actual events, for sometimes pleasant things could be punishments and adversity could be a reward. Others, however, did not shrink from pointing to particular misfortunes as divine judgments, as though God altered the behavior of things or beasts in order to work some outcome for the Church or its enemies.
It is in this milieu that Francis Galton (1872) sought to test the efficacy of prayer.[4] Indeed Galton quoted W.F. Hook’s Church Dictionary (1842): the general providence of God acts through what are called the laws of nature. By his particular providence God interferes with those laws, and he has promised to interfere in behalf of those who pray in the name of Jesus.
This belief that special providence interferes in the workings of nature implies that there should be a measurable difference in outcomes due to prayer. Thus Galton hypothesizes that prayerful people ought to obtain the temporal objects they seek more often than materialistic people.
Galton’s hypothesis resembles the popular belief B1 more than any doctrine-based hypothesis. Indeed, he explicitly appeals to the popular teachings and practices of Protestants and Catholics, which evince a deep belief in the power of prayer to obtain temporal goods.
Testing the hypothesis by observation and statistical analysis entails comparison of outcomes, without any supposition about the natural or supernatural means whereby the outcome might be achieved. In fact, Galton supposes that prayer is made effective by some subtle divine influence on human actions, rather than by direct physical interference, but this supposition is not essential.
Galton posits several possible statistical studies that might be conducted, such as: (1) to see if prayerful people are healed more rapidly than then the non-prayerful; (2) if missionaries have greater immunity to disease; (3) if the business enterprises of pious people, or on their behalf, have greater likelihood of success. In a word, he thinks we should test whether pious people have greater temporal success than the impious.
This formulation is problematic with respect to Christian doctrine, for we have noted that God may grant the entreaties even of the impious, and that Christians should prefer spiritual goods over the temporal. Indeed, the saintliest Christians despise temporal goods to the point of wishing for martyrdom or at least renouncing worldly goods and embracing whatever cross is sent to them. Galton would test a more popular sort of religion, contaminated by worldly ambition.
Despite his confident expectation of null results in all proposed studies, Galton offers concrete data for only one measure, showing the longevity of persons of various classes. He notes that members of royal houses are shorter lived, though nearly all of England prays for their long life. Clergy indeed are among the longest lived, but Galton thinks the difference is too small, and anyway easily explained by their easy country life. He notes that the more eminent clergy have shorter lifespans than eminent lawyers. Even supposing that clergy were so base as to pray for their own longevity, Galton’s use of data is biased. When it does not give the result he wishes for clergy, he chooses a subset that gives him a favorable result, and he explains away the undesired result. This selection bias is the main weakness of observational studies.
Galton’s challenge was not directly taken up by scientists in the form he presented. There continued to be many case studies of miraculous healings, most notably by the Bureau des Constatations at Lourdes, but no attempt to test the large scale efficacy of ordinary prayer. Nonetheless, as we shall see later, there were in fact many epidemiological studies in Europe and North America that touched on the effect of piety and religion on health, at least obliquely.
In 1954, the Commission on Divine Healing, established by the Archbishops of Canterbury and York, asked the British Medical Association to prepare a statement on the evidence of the physical or psychological value, positive or negative, of spiritual healing services and of private and public prayer. The BMA assigned this task to six doctors who were representatives on the Churches’s Council of Healing. Three Catholic doctors were also invited, but these all declined. Nine more doctors participated in discussions with the committee, another ten submitted memoranda, and another seventy responded to questionnaires. A summary of findings was published in a supplement to the British Medical Journal in 1956.[5]
The committee received numerous case histories of recoveries from apparently incurable ailments. Some of these, in the view of the committee, were attributable to mistaken diagnosis or prognosis. Others may be attributable to remission of symptoms such as pain while the malady remains. It was admitted that sometimes cures from deadly disease occur spontaneously for no known reason. In such cases, anything, including spiritual healing, might be the cause. In other cases, medical treatment was also given, so that might have been a contributing cause.
When all these possibilities are considered, it leaves little room for miraculous cures of organic disease by the methods of spiritual healing. In any event, spontaneous or unexpected cures in this country, like those of Lourdes, which cannot be explained are very few; and in the Committee’s opinion it is probably better to acknowledge that they are at present inexplicable on scientific grounds.
Here the committee appears to consider only properly supernatural miracles as a possible mode of divine healing. These are indeed few. Anything that admits a possible naturalistic modality, even in principle, is denied to be an effect of prayer, neglecting to consider, as Galton did, that prayer might be rendered effective through special disposition of natural agents, possibly resulting in measurably disparate outcome frequencies. This blind spot accounts for why the committee says:
As far then as our observation and investigation have gone, we have seen no evidence that there is any special type of illness cured solely by spiritual healing which cannot be cured by medical methods which do not involved such claims. The cases claimed as cures of a miraculous nature present no features of a unique and unexpected character outside the knowledge of any experienced physician or psychiatrist.
Here the committee is merely saying that all of the types of afflictions ostensibly cured by spiritual healings can sometimes also be cured by natural medicine. That is, there are no known maladies curable solely by spiritual healing, at least not among the cases investigated by the committee. Again, this is imposing a highly restrictive scope on the potential efficacy of prayer, as though divine healing should apply only to types of maladies that are categorically untreatable by natural medicine.
The committee attempted no statistical analysis that might show outcome frequency for various common illnesses, as its study seemed restricted to miraculous
cases. It provided uncontradicted reports of spontaneous cures of grave illnesses, usually after some sort of spiritual healing service. In one case, the patient was not known to have attended any such service, though she was a devout Christian and quite plausibly the recipient of some prayer.
Although the committee had a mostly positive view of the psychological effect of prayer on a patient’s hopefulness and well-being, which may improve medical outcomes, it denied that any direct physiological effect of prayer was proven. Spontaneous cures, they said, might be attributable to some unknown natural agency or to chance. Such a statement could be made about any unexplained phenomenon, so this position amounts to an a priori denial of the demonstrability of miraculous cures. It does not even consider the possibility that prayer may be no less truly effective even in non-miraculous cases.
Galton’s original program of retrospective observational investigation using statistical analysis remained relatively neglected. As we shall see later, there were in fact many observational health studies in the late nineteenth and early twentieth centuries that collected religious data about patients, enabling comparison of outcomes. Many of these studies were not even concerned about religion or prayer, but happened to collect relevant data. Thus there was no recognized body of scientific literature on the subject until the 1960s.
Over the past half-century, fewer than a hundred seriously scientific experiments on the efficacy of prayer have been conducted, and most of these have such severe methodological flaws as to be not worth considering. Among the few dozen that remain, there is a disparity of results, ranging from positive effects of prayer to no effect to even a slight negative effect. This disagreement need not imply that one or another study is invalid, for we have no reason to suppose that prayer should be uniformly efficacious across place and time, especially if it is subjected to the artificial constraints demanded by study design.
We have noted that all experimental studies, insofar as they put God to the test,
might be considered a form of divination, from which we should expect no positive result. Even if God should choose to honor some prayers regardless, it is unethical and impractical to impose strict controls, so such studies would be invalid. Apart from these unavoidable methodological shortcomings, various studies have shown additional weaknesses.
Possibly the first attempt to prove the power of prayer by scientific experiment and statistical analysis of outcomes was Rev. Franklin Loehr’s book The Power of Prayer on Plants (1959).[6] Loehr claimed to prove that seeds were more likely to sprout and grow faster if aided by prayer, and more controversially, that they could stop growing or be killed by negative prayer or curse. While the subject of praying for plants seems frivolous, it has the advantage of removing any placebo effect, insofar as plants are presumed to be non-sentient so they cannot know they are receiving prayer.
The first known double-blind clinical trial on humans testing prayer efficacy was published by C.R.B. Joyce and R.M.C. Welldon (1965).[7] Thirty-eight out of forty-eight patients were selected and paired with patients who could be matched as closely as possible for age, sex, diagnosis, and day of visit. One member of each pair was randomly allocated to the prayer treatment
condition by coin toss. All patients continued to receive medical treatment. Twenty-eight subjects were Anglican, three were Catholic, six were Jewish, and one was agnostic. Five of the intercessory prayer groups were organized by a non-denominational Christian body, and a sixth was organized by Quakers. Each group was given only the first name and a fictitious surname initial of each subject. Each group was allowed to determine its own mode of intercession, though they could not meet the patient, nor were they given follow-up information on the patient’s status. In all cases, they chose the practice of the presence of God,
a silent meditation focusing on some phrase of Scripture, while thinking of the patient in the context of the love of God. This is akin to what more recently has been called centering prayer.
The outcomes measured were the patient’s clinical state and attitude toward the illness (rheumatic or psychological maladay with poor prognosis). Each patient’s net change was compared with his partner’s. Clinical state was measured on a five-point scale, and attitude on a three-point scale. Two of the nineteen pairs were eliminated because they were later found not to have met the criteria for admission. (Such exclusion is methodologically correct.) Had they been included, there would be two more pairs with an advantage in clinical state associated with prayer. A third pair had to be eliminated because one of the patients refused to attend the follow-up, leaving only sixteen pairs to be studied.
Comparing patients treated
by prayer against controls, without any regard for pairing, the treated group had clinical improvements in 5 of 16 cases, while the control group had only 1 improvement in 16 cases. This difference is significant to 8 percent by Fisher’s exact test (i.e., only 8 percent of random samples of patients of this size would give such a large difference). If we compare members of each pair, however, we find less disparity, as the treated patient had a better net change than his partner in seven pairs, but the controls did better in five pairs. Some of this is due to accidents of pairing, but also from the fact that there were more cases with negative changes among treated (7) than controls (5). The attitude data was even less conclusive, due to the limited scale of just three different values.
A triple-blind study conducted by P.J. Collipp (1969) studied 18 children with leukemia, 10 of whom were randomly selected for intercession by a prayer group in another city.[8] No attempt was made to match conditions in groups. After 15 months, 7 of 10 children in the prayer group survived, while only 2 of 9 children in the control group survived. I confirmed by t-test Collipp’s claim that the result is significant at 0.10 (i.e., only 10 percent of similarly sized samples would yield such disparity). Rosner (1975) complained that this difference is attributable in part to inequity in condition (two controls had acute myelogenous leukemia, which has poorer prognosis) and variety in age and treatment.[9] Yet these variations are randomly distributed between samples, so such inequity is already built into the statistical significance calculation. In both of the 1960s studies, we find statistically significant results (taking into account the small sample sizes) in the aggregate, though not in paired or matched comparisons.
The most influential study on prayer was that conducted by Randolph Byrd (1988).[10] In 1982-83, 393
coronary care patients were enrolled in a double-blind clinical study of intercessory prayer. Another 57 were eligible but declined to participate. Intercessors were born again
Christians (including some Catholics) with active church membership and daily devotional prayer. Intercessors were given the patients’ first name, diagnosis, and updates in their condition. This last had not been done in previous studies, for fear it would alter the behavior of intercessors. Although patients were randomly assigned to intercessory prayer (N = 192) or control (N = 201), Byrd took care to check that the two groups had no statistically significant difference in severity of illness.
The study used multivariate analysis, as it measured 29 therapeutic outcomes or events in each group. For 23 of these outcomes, there was no statistically significant difference at a level of 0.05. This included days in the cardiac care unit, days in the hospital, number of discharge medications, mortality (13 prayer to 17 control), pressure monitoring, and various other interventions. In many cases, the number of incidents was too small in both groups to measure any significant difference.
For six variables, there was a difference with statistical significance p < 0.01, i.e., 99 percent rather than the usual 95 percent. All six of these variables showed better outcomes for the prayer sample. These were: congestive heart failure (8 prayer to 20 control), diuretics (5 to 15), cardiopulmonary arrest, (3 to 14), pneumonia (3 to 13), antibiotics (3 to 17), intubation/ventilation (0 to 12). As Byrd himself notes, however, these six variables do not suffice to establish statistical significance for the study, since there were many variables examined, and one might expect findings for some variables by chance.
To test whether the overall results for the 29 variables were statistically significant, Byrd performed a multivariate analysis, revealing a significant difference on the level of p < 0.0001. Additionally, he examined overall scoring of patient outcomes as good, intermediate or bad (based on net increase in morbidity or risk of death), which effectively combined all variables into a single outcome. This showed a statistically significant effect at p < 0.01.
Byrd acknowledged that even the control group may have received prayer, but this is ethically and practically unavoidable, and in any case may have resulted in smaller differences observed between the two groups.
This assumes (1) that the unregulated
prayer was received more or less equally in both randomized groups and (2) that the efficacy of prayer is in some way proportionate to quantity. Both claims are unproven, but perhaps they are unavoidable assumptions of any intercessory prayer study.
The positive result of the Byrd study caused a sensation, notwithstanding criticisms of Byrd’s religious bias and his methodology of looking for correlations among large numbers of variables. Naturally, there was motivation to see if these results were replicated in other studies.
Harris et al. (1999) randomized 990 coronary care unit patients at the Mid America Heart Institute (Kansas City) into a control group (N=529 after 5 were excluded for CCU admission less than 24 hours) and prayer treatment group (N=466 after 18 excluded).[11] Staff and patients were blinded not only to condition, but to the very existence of the trial. The institutional review board had determined, somewhat controversially, that the study did not require informed consent, as there is no potential harm in intercessory prayer. Prior to the study, cardiologists developed a weighted scoring system of clinical outcome variables to measure overall quality of outcome, with a lower scores representing a better outcome. Fifteen teams with five intercessors, all of whom attested that they believed in a personal God who answers prayer, were given each patient’s first name. They were asked to pray for a quick recovery with no complications, and anything else they wished to add. With this design, the expectation was that any observed effect should be in the overall outcome score, rather than in any particular clinical outcome variable.
The result was that patients in the prayer condition had an 11% lower (better) score than the control group. This was significant to p < 0.04, so that there was only a 1 in 25 chance of such a study achieving this result by chance if there were in fact no effect correlated to prayer. For comparison, they also measured patient outcomes by Byrd’s simplified (good, bad, intermediate) scale, but this showed no statistically significant difference.
The following year, several reviews of prayer studies conducted to date were published. Roberts, Ahmed and Hall (2000) noted that there was little difference in death rates and little change in heart problems with bad or intermediate outcome, though prayer increased chances of not being readdmitted to CCU.[12] They considered it too soon to tell if there was a definite effect of prayer. Abbot (2000) reviewed 22 trials, 10 of which reported significant effects of spiritual healing. Only 8 studies had a maximum methodological quality score, and 5 of these showed a significant outcome for healing.[13] Apart from the two large-scale trials mentioned (Byrd and Harris), most prayer trials were too varied in method to enable pooling of results to draw global conclusions. Likewise Astin et al. (2000) reviewed 23 trials and found they were too dissimilar for meta-analysis. 57% of these trials showed positive treatment effects, so further research was merited.[14]
In 2001, there were three published studies of note. The oddest was that of Leibovici, who, with tongue firmly in cheek, tested the hypothesis that prayer might be retroactively effective, as the intercessors prayed in 2000 for hospital patients with bloodstream infections in 1990-96. An advantage of this design was the ability to obtain large samples: 1691 patients in the intervention group and 1702 in the control group. Curiously, all three measures showed favorable differences for the intervention group, two with statistical significance at p < 0.05. There was a small reduction in mortality (28.1% vs. 30.2%) and significant reduction in length of hospital stay (p=0.01) and duration of fever p=0.04.[15] The logically problematic hypothesis should give us pause, for statistically significant correlations can arise even in the absence of causation. Indeed, the point of Leibovici’s study was to show how problematic it is to test
an effect without a well-defined causal mechanism.
Similar caution is advised in the far more remarkable study by Cha, Wirth and Lobo (2001), in which intercessory prayer reportedly improved in vitro fertilization (IVF) success rates in Korean patients.[16] Pregnancy rates increased from 26% to 50% (p=.0013), and implantation rates increased from 8% to 16.3% (p=.0005). The astoundingly high success rate was unprecedented for any intercessory prayer study. Doubt was cast on its results when coauthor Daniel Wirth pled guilty to mail and bank fraud on an unrelated matter in 2004. Dr. Cha nonetheless has defended the study’s integrity, insisting that it was impossible for Wirth to have interfered in the conduct or outcome of the clinical trial. For Catholics, this finding is problematic, as it would indicate that God should grant special favor to an act of mortal sin.
The study had an odd design, in that a second tier of intercessors prayed for the effectiveness of the prayers of the first tier intercessors. This supposedly would magnify the overall effect. Another three individuals prayed in general that God’s will be done for the intercessors in the first two tiers, ostensibly to avoid any charge of putting God to the test or presumption. There was certainly no placebo effect, as the subjects were not even aware they were in a study, much less of their condition of intervention or control. Despite the strong results, the authors cautioned that we view these data to be preliminary and that they may not be confirmed in future investigations.
That same year saw yet another coronary care trial, akin to those of Byrd (1988) and Harris (1999). Aviles et al. (2001), unlike the earlier studies, found no significant effect, though the non-significant effects were consistently in the direction favoring the intercessory prayer group. There was at least one adverse event in 25.6% of the intervention group (N=400) and in 29.3% of the control group (N=349). 31% of intervention group patients had primary outcomes vs. 33.33% of controls. In low risk patients, the incidence of primary outcomes was 17% for the intervention group and 24.1% for controls.[17]
Another cardiac study, called MANTRA II, published in the Lancet in 2005, examined two distinct noetic
interventions: (1) intercessory prayer and (2) music, imagery and touch (MIT) therapy. 748 patients were subjected to 2 × 2 randomization, resulting in groupings of: 371 prayer; 377 no prayer; 374 MIT therapy; 374 no MIT therapy. Cross-grouping numbers were 192 standard care only; 182 prayer only; 185 MIT therapy only; 189 both prayer and MIT therapy. Patients were blinded to whether they received intercessory prayer, but such blinding was not possible for MIT therapy. Primary and secondary outcomes to be tested were prespecified. The primary outcome was a combnination of in-hospital major adverse cardiovascular events and readmission or death after 6 months. The three secondary outcomes were: major adverse cardiovascular events; readmission or death after 6 months; mortality after 6 months. Krucoff et al. (2005) reported no significant difference for the primary composite outcome in any treatment comparison. However, the secondary endpoint of mortality was lower with MIT therapy than without MIT therapy, yielding a hazard ratio of 0.35, with 95% confidence interval of 0.15-0.82.[18]
The positive secondary outcome for MIT therapy might be attributed at least in part to a placebo effect, since subjects could not be blinded. The null result for intercessory prayer means that any positive effect, if it existed, was too infrequent to be statistically significant. The slightly beneficial hazard ratios (< 1) for intercessory prayer were consistent with no effect (≥ 1).
Primary outcome: 0.97 (CI 0.77–1.24)
Secondary outcomes Readmission or death after 6 mo.: 0.93 (CI 0.72–1.19)
Major cardiovascular event: 0.85 (CI 0.63–1.14) 0.2785
Mortality after 6 mo.: 1.13 (CI 0.53–2.4)
Mortality was much too infrequent even in controls to enable detection of any effects, but significant effects might have been measured for other outcomes, yet they were not. An earlier pilot (MANTRA) in 2001 had studied 4 noetic therapies, among which intercessory prayer had the lowest complication rates. This study was only for determining feasibility. At that time, the authors determined that, to measure a 30% treatment effect, they would need 600 patients in each group to have an 80% likelihood of being able to detect such an effect with 95% certainty.[19] In fact, the MANTRA II study would have fewer than 400 subjects per group, so it was only powered to measure effects much larger than 30% as significant.
The largest and most widely publicized study of intercessory prayer in cardiac patients was published in 2006. The STEP study by Benson et al. tested both subjective and objective effects of intercessory prayer using a three-group design in cardiac bypass patients: (1) intercessory prayer for subjects blinded to condition; (2) control subjects blinded to condition; (3) intercessory prayer for subjects informed of intervention. The blinded subjects were aware of the study and that they may or may not be prayed for.
[20] Assuming that 50% of controls would have a complication and that complication rates of 40% and 30% in Groups 1 and 3 would be important effects, the investigators arrived at a desired sample size of 600 per group to detect this effect. Actual sample sizes were 604, 597, and 601 for Groups 1-3. The primary outcome to be tested was all-or-nothing: any post-operative complication within 30 days of bypass, or none. Secondary outcomes were the presence of any major event and mortality.
The intercessors (consisting of two Catholic monastic groups and one Protestant group) were asked to add a specific phrase to their prayers: for a successful surgery with a quick, healthy recovery and no complications.
While the investigators understandably wanted a clear hypothesis to test, a prayer for absolutely no complications borders on the frivolous. Most isolated complications (except for renal failure) are not associated with an increased risk of death.[21] This prayer is not so much concerned with the safety and health of the patient but with putting God to a test. Admittedly, the same might be said of any controlled study of intercessory prayer. Another methodological shortcoming, common to all such studies, is that practically all subjects in all groups believed they had friends, relatives or church members praying for them (574/604 in Group 1, 579/597 in Group 2, 577/601 in Group 3).
The results were null for the effect of intercessory prayer (Group 1 vs. Group 2) for primary and secondary outcomes. Although these were all-or-nothing outcomes, the same null result is found even if one were to treat specific kinds of complications as distinct variables. Interestingly, Group 3 patients had a statistically significant increased risk of complication (relative risk 1.14, 95% CI 1.02-1.28) vs. Group 1, as though knowing for certain that they were being prayed for had some negative effect. The strongest complication increase was in new onset atrial fibrillation/flutter. There was no statistically significant increase in risk for major events or mortality.
The decisively null result of the STEP trial on the efficacy of intercessory prayer led opponents of religion to declare that this was a definitive test, and that all previous experiments showing positive results were methodologically flawed. Defenders of religion, by contrast, pointed out that the STEP study could not control for private prayer and that its contrived intercessory prayer was insincere and presumptuous. Both sets of criticisms are misguided. The Byrd (1988) and Harris (1999) studies were of high methodological quality, and it would be unscientific cherry-picking to discard or downgrade studies on the basis of their results. The flaws of the STEP study, for its part, were not markedly different from those common to all interventional studies of intercessory prayer.
Recent studies on non-cardiac physical health outcomes include that by Astin et al. (2006) on AIDS patients and that by Rosa et al. (2013) on pregnancies. Astin (2006) compared two intercessory groups, one with professional
intercessors and one with nurses acting as intercessors, against a control group (about 40 in each group). There were no significant effects by either intercession for pre-defined primary or secondary outcomes. There were positive outcomes, however, on two variables: 6-month reduction in CD4 + lymphocyte counts, and 12-month triglyceride levels. Interestingly, the subjects receiving intercession from either group were significantly more likely to guess that they had been receiving healing
than the control subjects.[22]
Brazilian researchers Rosa et al. (2013) studied 564 pregnant women. 281 of these received intercessory prayer (blind to condition) for nine days from a group of six women, coordinated by a theologian, asking for good delivery and health of the newborn. Both randomized groups had similar backgrounds and gestational age distribution, though the controls were significantly more likely to have had previous abortions (18.3% vs. 10.3% of prayer group). There was no significant difference in any of the serious adverse events measured: low Agpar score, preterm birth, low birth weight (10th percentile), high birth weight (> 4 kg), caesarean delivery, spontaneous abortion, intrauterine death.[23] It should be noted, however, that all adverse outcomes except caesareans were infrequent in both groups (15 or fewer patients per group), making the detection of an effect less likely.
Several intercessory prayer studies have also been conducted on psychological outcomes. Walker et al. (1997) studied 40 patients treated for alcoholism in a double-blind study, assessing outcomes after 3 months and 6 months. No difference was found in alcohol consumption.[24] Overall the subjects (intervention and control combined) had a delay in drinking reduction compared to a normative group of subjects in same facility. Those aware of a family member (not a study intercessor) praying for them at baseline were found drinking significantly more at baseline than those not so aware. (This correlation need not be causal; it could just be that harder
cases are more likely to attract such concern from family.) The study found that greater frequency of prayer by the participants themselves was associated with less drinking, but only for months 2 and 3. The authors found the results to suggest that the subjective effects of prayer may involve complex interactions.
In 2004, Palmer et al. published a controlled, blinded study of 86 Presbyterian churchgoers (45 intervention and 41 control). Each subject disclosed a specific physical or psychological problem. Subjects in the intervention group did not know they were receiving intercessory prayer. The 12 intercessors were volunteers, who were free to pray as often as they wished (at least once a day) for whatever duration, with or without a script.
A standardized medical outcomes scale was applied to assess improvement. On that basis, no intervention effect was found. However, the study did find a marginally significant reduction in the amount of pain
in the intervention group. Also, those who believed that their problem could be resolved by prayer had significantly lower concern about their problem at the end of the study.[25]
Boelens et al. (2009) found that intercessory prayer significantly improved clinical anxiety and depression compared to controls. They also found that intervention subjects maintained more daily spiritual experiences and optimism one month after intervention.[26] In this study, however, the minister was often with the patient, so the subject was not blinded to condition, and this effect cannot be separated from a placebo effect or the subjective psychological benefit of prayer.
Dehghani et al. (2012) studied the psychological benefit of prayer in Iranian mothers of children with cancer. The sample size was powered to test an effect of 80% at p < 0.05. Half of the 60 women studied were taught prayer therapy. This was not praying for a particular outcome, but conventional Muslim practices of praying 3 times a day for 10 minutes, preferably in mosques, reading the Quran, etc. They were given Speielberger’s State Anxiety Inventory before and 21 days after the intervention. Anxiety scores in the intervention group were lowered by more than a full standard deviation, significant at p < 0.001.[27] This strong indication of the psychological benefit of the prayer may suffice to account for the ubiquity of the practice, even if evidence of so-called objective
effects is more equivocal. After all, the psyche is no less real than the body, and prayer is primarily oriented toward spiritual benefits.
A literature review by Simão et al. (2016) found only 12 out of 92 scientific papers to report unique randomized clinical trials on patients with prayer as the intervention, ordinary medical care as the control, and any health change as an outcome. Of the 12 papers analyzed (all of which we have discussed), 11 scored 3 or higher in the methodological criteria of Jadad (1996),[28] with the exception of the Boelens study, where patient blinding was not possible. They found that 7 of these 12 studies considered prayer to be a positive factor, reducing anxiety or improving physical functioning in those who believe in prayer.[29] The authors state that all 10 intercessory prayer studies (excluding Boelens and another unspecified) applied prayer over a period less than one month. This is incorrect; Joyce (1965) had prayer for 6 months. By focusing on the subjective benefits of prayer, the review evaded discussion of the objective benefits, the reality and measurability of which have been controversial.
While there have been barely a dozen high quality randomized interventional studies of prayer, with apparently conflicting results, by contrast there have been hundreds of epidemiological observational studies of the effect of religion on health, and they have consistently shown measurable salutary effects of religion. Some of these studies are retrospective, much as Galton proposed, while others are prospective, even longitudinal in design, following the same subjects over years or decades.
The vast majority of epidemiological studies on religion and health do not focus on prayer per se, but use church attendance or affiliation as a marker of religiosity, to compare outcomes with those who are less religious. Presumably, those who attend religious services more frequently will be more often inclined to prayer. Many studies have found statistically significant correlations between such religious indicators and positive health outcomes. This correlation is not proof of causality, however.
Sound studies control for prevalent health conditions upon enrollment. Otherwise, the effect of religiosity or prayer may be confounded with the effect of salutary practices such as abstinence from smoking or drinking. It is unclear, however, how far one should go in controlling for health variables. In most studies, people already have their established religious beliefs and practices upon enrollment, and their better behavior and health conditions may be in part a result of their religion. If we exclude the benefits of these behavioral choices, we are limiting our study only to involuntary effects of religion. If we go further and control for every measurable health variable, down to lab results for every chemical, we would effectively restrict our search to purely supernatural effects, which are rare. The choice to control for all health variables is especially odd in comparative studies of mortality. It is as if we should expect God to lower the mortality of the pious without also improving their health (for the subjects were religious prior to enrollment).
The consistent findings of a positive association between religious service attendance and lower mortality need not imply a finding for the efficacy of prayer. Even after controlling for health variables, the health benefit of religious participation might be driven by the psychological benefits of social interaction, rather than a result of frequent petitionary prayer. In fact, Catholics ought to expect the most pious to be least likely to pray for their own longevity. One is hard pressed to find any saint making such a petition; on the contrary, many prayed for their life to be cut short by martyrdom or some other form of self-sacrifice.
Levin and Schiller (1987) attempted the first comprehensive review of epidemiological literature on religion and health. Prior to this, there was little awareness that such a literature even existed, mainly because it consisted of clusters of articles on diverse health topics, where religious attendance or affiliation were variables, but not always the main topic of interest. B.H. Kaplan (1976), among others, had reviewed the literature on heart disease,[30] and later contributed to a study showing that frequent church attenders had lower blood pressures even after controlling for effects of age, obesity, smoking and socioeconomic status. [31]
Levin and Schiller cast a much wider net, finding well over 200 articles, mostly in European and North American journals, that dealt with religion and health variables. The preponderance of these studies dealt with religious affiliation (Catholic, Jewish, Protestant, Mormon, Muslim, Zoroastrian) rather than church attendance as a variable. Catholics had lower coronary artery disease than Protestants, who fared better than Jews. Italian Catholics had lower coronary heart disease mortality than New York Jews, despite having higher cholesterol. Some of the positive health outcomes, such as lower cancer incidence in reproductive organs, might be attributable to more salutary behaviors, yet most studies showed little interest in examining religion as a factor in health behavior. Two studies of Catholic priests showed lower mortality, which the authors conjecturally attributed to high role satisfaction. Higher mortality among missionaries was assumed to be linked to higher exposure to disease. While the literature overwhelmingly favored a positive link between religion and many health variables, the mode of causality was ill-defined.[32]
In a 1994 review, Levin re-asssessed the epidemiological literature, including recent literature reviews by other authors, to ascertain evidence of a causal association between religion and health. He found that there is substantial empirical evidence for an association between religion and health,
established in hundreds of studies of scores of diseases, finding various religious indicators (affiliation, frequency of attendance, level of priesthood) to be positively correlated with salutary outcomes. This association is probably
valid in the sense of not being due to chance, bias or confounding variables. While it is certainly possible that the positive health effects of religion come about through the behavioral changes it promotes, these would nonetheless be genuine religious effects, though perhaps not supernatural effects. We cannot eliminate the possibility that there might be some confounding variable that is not a function of religion, if only because observational studies, by their nature, do not rule these out. The question of causation is most difficult of all, not only for religion but all epidemiological studies. Using Hill’s verificationist criteria supporting causal hypotheses, Levin found the evidence of a causal relation between religion and health to be inconclusive, but promising.
[33]
Since that time, there have been further studies confirming that higher degrees of religious involvement have positive correlations to low mortality and other salutary outcomes. The magnitude of the effect is striking: in one study of 22,080 people, those with no religious attendance had 1.87 times the mortality risk of those who attended services more than once per week. A significant effect of 1.50 remained even after controlling for socioeconomic and health behavioral variables.[34] As others have cautioned, however, these studies of the effect of church attendance do not translate into the effects of specific practices such as prayer or Bible reading.[35] A few studies have attempted to identify effects of specific religious activities, such as one which found lower blood pressures among those who prayed or studied the bible daily.[36]
A recent longitudinal study (2005-2013) of black women examined frequency of prayer as a variable with respect to mortality. This study found the usual strong correlation between frequent religious attendance and lower mortality rate (0.64, 95% CI 0.51-0.80) compared to non-attendances, even after controlling for demographic and health variables. Strikingly, however, the reported frequency of prayer had a negative correlation to lower mortality, though falling short of statistical significance. Those who prayed serveral times a day had a mortality ratio of 1.28 (95% CI 0.99-1.67) compared to those who prayed less then once a week.[37] Oddly, there were mild positive (non-significant) effects for those who claimed to be very religious or spiritual, and involved religion in dealing with stressful situations. This suggests that frequency of prayer might not be a good measure of piety, if such a measure is even possible. More importantly, this study shows that most of the positive health effects of church attendance are not mediated by private religious practices.
A systematic review of 241 studies over 30 years 1988-2017 on the psychological effects of religion in adolescence found that religion is generally adaptive for adolescents, protecting them from risk behaviors and mental illness, and promoting youth development. Some evidence in experimental and longitudinal studies suggests this relation is causal.[38] That indication concurs with an earlier study in 2,676 adults (tracked 1965-1994), aged 17-65 at enrollment. That study found that those with weekly religious attendance were more likely to improve and maintain good mental health, increase social relationships, and improve poor health behaviors. The association between religous attendance and mental health and behavior improvement was stronger in women, consistent with findings that women have a stronger survival effect from religious attendance.[39]
Victorian agnostics believed that the Church of England’s prayer for the sick provided an opportunity to test the efficacy of prayer empirically. Galton considered that a comparison of mortality between the religious and irreligious would indicate whether such common prayer was effective. Numerous epidemiological studies conducted over the past century have in fact consistently found a strong beneficial effect of religious affiliation and church attendance on all-causes mortality and various chronic diseases. Some of this benefit, but apparently not all of it, is mediated by religiously motivated improvements in health behaviors, social relationships, and mental health. We cannot say how much of this benefit is due to the common prayers of churches for the sick. Notably, however, there appears to be no correlation between longevity and personal prayer or devotion. This comports with what we know of the saints and other pious orthodox Christians, who prayed not for their own health or long life, but for the grace to pass through whatever trials God may send.
Ordinarily, the best way to determine whether correlations with health variables are causal is to conduct a randomized controlled clinical trial. When applied to prayer, however, this approach actually introduces methodological problems not found in observational studies. These include putting God to the test, using contrived forms of prayer for ends of dubious worth, and the falsity of controls due to personal or family prayer, making all prayer studies effectively a test of a dose-response paradigm of strictly intercessory prayer. Some studies appear to be merely fishing expeditions for correlations among health variables, with no prior expectation of where to find an effect. Such approaches can invalidate the statistical tests for significance. Most problematic of all, testing any definite hypothesis in science requires positing some definite mode of causality. Prayer studies and meta-analyses have largely shied away from defining how prayer might be effective, naturally (A5) or supernaturally (A4), temporally or atemporally (A6), though probabilistic analysis cannot validly ignore such considerations. Those who do articulate a mechanism seem to be testing prayer as a sort of magic, acting at a distance, as though the words and thoughts of the intercessor are the thing that has some remote therapeutic effect. To be sure, this matches some popular characterizations of prayer, but it is barely compatible with a doctrinally based conception.
Unsurprisingly, in view of the serious methodological problems outlined, the clinical trials of prayer have had equivocal results. These studies tested the efficacy of prayer in a purely correlative sense, without judging the mode of causality, and in that sense they tested the hypotheses we have called A3, B1, and B2. Among the fifteen clinical trials we examined (which were the best in methodological quality), seven found a significant positive effect, seven had null results, and one found a negative effect. Even here we must be careful, for three of the effects (two positive, one negative) were in groups of subjects who were aware of the prayer. Moreover, some of the null
studies had positive findings on some variables, while some of the positive studies had null results for some outcomes.
In the statistical design of a clinical trial, the number of subjects is informed by the size of the effect one expects or wishes to measure. Joyce and Welldon (1965) tested for a 35% cure rate, on the basis of Beecher’s (1955) finding that the placebo effect could have cure rates that high.[40] It is unclear why they should choose this threshold, especially when studying patients with poor prognosis, given that they were already controlling for the placebo effect, so the measured improvement rate (19%) would not be attributable to it. Moreover, Beecher’s finding has lately been challenged, as have the entire methodology of measuring the placebo effect and the effect’s existence.[41] The placebo effect, being ostensibly psychogenic, tends to be more pronounced in symptoms or diseases of the nervous system. A recent meta-analysis of antiepleptic agents showed that seizure frequency was reduced more than 50% in 9.3%-16.6% of placebo patients, which is 20%-50% of the effect of active agents.[42]
In clinical studies of prayer, all patients also received regular treatment, so any measured effect would be in addition to that of normal treatment. This can be expressed in the form of a hazard ratio, which is not the same as an absolute cure rate. For all we know, prayer would be effective
more frequently if medical treatment were unavailable. After all, the efficacy of prayer is determined not be some materialistic modality, by the sovereign Divine Will, who is free to grant or deny petitions. Contrary to some agnostic complaints, God is not bound in justice to heal based on the severity of the illness, and He may grant undeserved favor to whomever He pleases. The claim that this is unjust or unfair repeats an Origenist error, and is informed by modern liberal egalitarian ideology, which would abolish all privilege.
Since the efficacy
of prayer rests on the sovereign Divine Will, the contradictory
results of the clinical trials need not imply that one or another set of trials had false results. There is no obligation for God to grant prayers with the same frequency in all cases, so prayer may have been measurably efficacious in some studies and not so in others. Science ordinarily tests hypotheses about regular causality, for which an expectation of repeatibility is appropriate. Yet the supposition of regularity is inappropriate for prayer.
The lack of a definite causal mechanism makes it impossible to apply Bayesian analysis to prayer studies, for we cannot say what the prior probability is. Jørgensen et al. (2009) say the prior probability is small, based on their unargued philosophical assertions that the existence of God, the communicability of prayer to God, and the ability of God to respond to prayer are all unlikely to be true.
[43] The metaphysical weakness of their position, which includes the error that the wisdom of God precludes human attempts at petition, only emphasizes the inapplicability of the model they are trying to use.
Most clinical trials use frequentist probability, and apply statistical tests (usually the t-test) that involve assumptions such as random sampling of a population, variables following a normal distribution, and equal variance of compared samples. These are reasonable assumptions for many stochastic processes, but perhaps not for something effected by an unfathomable Intellect. Moreover, the measure of statistical significance requires prior definition of the hypothesis so that it is mutually exclusive with the null hypothesis. Here we face the problem of defining in advance how large an effect we should expect, and in which variables.
Rosa et al. (2013) gave a thoughtful discussion of the methodological problems facing even rigorous studies. The lack of a biological mechanism makes it difficult to select a relevant outcome. The desired duration and type of intercession is also difficult to evaluate. They did not recommend further studies under the randomized clinical trial model. Rather, studies using different methodological approaches are required.
We are now faced with a new form of under-questioned authority and evidence from well-designed and methodologically appraised RCTs. The evidence from RCTs is now prized, even when it is incapable of providing meaningful information, particularly when the underlying causal theory is inscrutable. The evidence obtained from experimental trials in support of remote IP provides us with an illustrative view that highlights systematic scientific blind spots in the institutions of EBM. At its core, medicine, including evidence-based medicine, is based on theory. Thus, EBM must cultivate greater capacity to address the crucial role of theory in both the generation and use of experimental evidence.
While it is possible that the efficacy of prayer is altogether unmeasurable via randomized clinical trials, skeptics may retort that, at least in principle, an effect might have been measured, but in fact it has not been measured, at least not with a convincing level of consistency. This non-confirmation or equivocal confirmation, it is said, should be considered evidence against Christian claims of the efficacy of prayer. This is fair enough, but against that it must be said that, by the very standards proposed by agnostics over a century ago, comparing mortality of the pious against the impious, a resounding positive effect has been measured consistently in numerous epidemiological studies. As Galton recognized that prayer for the sick might only sometimes be efficacious in the sense of affecting outcomes, he sought only a large-scale population effect, which has indeed been measured. When we try to narrow down the source of this effect to something like the frequency of private prayer or the presence of intercessory prayer, however, the correlation mostly or completely dissipates, at least in some studies. Perhaps this failure should be unsurprising, since it supposes a quantitative dose-response model, i.e., that effectiveness derives from the frequency of personal prayer or from the additive effect of intercessory prayer. Yet Christ teaches no such thing about prayer, on the contrary disparaging quantity of words as a measure of effectiveness. (Mt. 6:7) Prayer is an offering of the heart to God, and it is only by this measure
that a prayer might be valued, though even here God has freedom in His inscrutable judgment.
Catholics will note that the confinement of this discussion to statistical effects in populations and clinical trial studies entails omitting a vast literature on miraculous healings. Such literature deals with case studies, which are analyzed forensically rather than by statistical methods. Since miraculous healings are infrequent occurrences, we do not expect them to be measurable in population studies or clinical trials. Nonetheless, many of these cases have been well documented and investigated, and include cures that are categorically impossible to heal by natural means, even regeneration. Philosophy has moved well beyond the need to refute the well-worn metaphysical errors that miracles are impossible since they would violate the laws of physics or else require new laws to explain them. The metaphysical contingency of the natural order is now well understood even by physicists. While the reality of miraculous healings is difficult to dispute, for anyone who has explored this literature in depth, we cannot use case studies to prove that such healings were caused by prayer, in the ordinary manner by which causation is proved, though causation is here a most reasonable inference.
In what sense, if any, are Christian claims about the efficacy of prayer actually testable, not merely in principle, but in practicable studies? If we restrict ourselves to doctrinally based hypotheses, we find that testing A3, the efficacy of intercessory prayer, is greatly complicated in practice by the methodological difficulties discussed. The contradictory results of the intercessory prayer studies are consistent with the truth of A3, for we need not expect such prayer to have temporal effects with regularity. Among the methodological problems is an inability to distinguish among hypotheses A4, A5, A6, defining the modality by which prayer is effective. The popular hypothesis B1, with its much stronger claim of efficacy, is more difficult, though not impossible, to reconcile with the conflicting results. Hypothesis B2, the effect depending on the number of petitions has not been directly tested, though such a study design is feasible. Insofar as most prayer studies have tacitly or explicitly assumed an additive effect of intercessory prayer, they may be said to have indirectly studied B2, again with conflicting results. Epidemiological studies comparing the religious and non-religious were once thought to test B2, on the assumption that only the religious would benefit from the prayers of the Christian churches. Yet Christians pray for all people, and God may heed the prayers even of non-Christians (A7).
Historically, Christians have believed in the efficacy of prayer on the evidence of specific cases, starting from the miraculous healings in the Gospels through those of modern times. Though many other kinds of miracles are possible, those of physical healing have held pride of place, owing to their exemplification of divine mercy and the Messianic mission of curing the person in his integrity, body and soul. Only in recent centuries has a new mode of evidence become possible, which might enable us to detect, on a population scale, the overall effectiveness of prayer in altering outcomes, even for ailments that sometimes admit natural healing, by comparing prayer intervention (or some surrogate such as religious participation) against controls. Christians, far from shrinking from this challenge, boldly faced it and found a wealth of evidence positively correlating religious practice and health variables, even in well controlled studies, contrary to the expectations of agnostics. Moreover, contrary to the claims of some atheists who regard religion as a mental illness, empirical evidence consistently shows positive mental health effects of religious activity. As is often case, it is the so-called skeptics who make uncritical assertions and then refuse to look at evidence bearing on their claims.
This evidence falls short, however, of definitive confirmation of the objective efficacy of prayer. We must agree with Rosa et al. that randomized clinical trials are not an apt mechanism for confirming or disconfirming this efficacy, due to the difficulty of setting the possible modes of causality within any definite constraints. The studies with positive findings are at most suggestive of objective efficacy, but without a mode of causality this remains an effect
only in a statistical sense.
Further study is needed, not by multiplying clinical trials, but by exploring the notions of causality implied by Providential action in honor of prayer, and positing appropriate methods of observation and models of probability corresponding to such notions. A similar discussion has arisen in physics, where the statistical interpretation of quantum mechanics and the underlying probability theory appear to be in need of reconceptualization in order to arrive at deeper notions of causality and temporality than those that described Newtonian mechanics. The exact modality of causation, and indeed whether causality is an apt description at all, remains a matter of controversy, in physics no less than in acts of Providence.
[1] St. Augustine. Tractates on the Gospel of John, 102 (Jn. 16:23-28).
[2] St. Thomas Aquinas. Summa Theologiae, II-ii, 83, 2.
[3] Ibid., II-ii, 83, 16. St. Thomas’s second condition, things necessary for salvation,
appears much more restrictive than St. Augustine’s things not contrary to salvation.
See, however, his discussion in II-ii, 83, 15, ad 2, which includes that which is useful
(utile) or conducive
(pertinens) to salvation.
[4] Galton, Francis. Statistical Inquiries into the Efficacy of Prayer.
The Fortnightly Review LXVIII, Aug. 1, 1872. Galton was responding to the so-called Prayer-Gauge debate, ignited by an anonymous essay submitted by John Tyndall (likely authored or inspired by the surgeon Sir Henry Thompson): Prayer for the Sick: Hints towards a Serious Attempt to Estimate Its Value.
London Contemporary Review, July 1872. This proposed an experiment comparing outcomes between two hospital wards, where both wards would receive the churches’ general prayer for the sick, but only one would receive special intention prayers of a congregation. This is the model of later intercessory prayer studies, but the controlled design is still frustrated, since one cannot ethically prevent friends or relatives from praying especially for a patient.
[5] Divine Healing: B.M.A. Evidence to Archbishops’s Commission.
Supplement to the British Medical Journal, Vol 1.,, No. 4975, May 12, 1956, pp. 269-273.
[6] Loehr, Franklin. The Power of Prayer on Plants Research Association Press, 1959.
[7] Joyce CRB, Welldon RMC. The Objective Efficacy of Prayer: A Double-Blind Clinical Trial.
J. chron. Dis. 1965, Vol. 18, pp. 367-377.
[8] Collipp PJ. The Efficacy of Prayer: A Triple Blind Study.
Med Time 1969, 97:201-204.
[9] Rosner, Fred. The Efficacy of Prayer: Scientific vs. Religious Evidence.
J Religion and Health, Oct. 1975, 14(4):294-298.
[10]
Byrd RC. Positive therapeutic effects of intercessory prayer in a coronary care unit population.
South Med J. 1988; 81(7):826-829.
[11]
Harris WS, Gowda M, Kolb JW, Strychacz CP, Vacek JL, Jones PG, Forker A, O’Keefe JH, McCallister BD. A randomized controlled trial of
the effects of remote intercessory prayer on outcomes in patients admitted to the coronary care unit.
Archives International Medicine 1999, 159: 2273–78.
[12]
Roberts L, Ahmed I, Hall S. Intercessory prayer for the alleviation of ill health.
[13]
Abbot NC. Healing as a therapy for human disease: a systematic review.
J Altern Complement Med. 2000; 6(2):159-69.
[14]
Astin JA, Harkness E, Ernst E. The efficacy of distant healing: A systematic review of
randomized trials.
Annals of Internal Medicine. 2000, 6:903–10.
[15]
Leibovici L. Effects of remote, retroactive intercessory prayer on outcomes in patients with
bloodstream infection: Randomised controlled trial.
British Medical Journal 323 (2001): 1450–51.
[16]
Cha KY, Wirth DP, Lobo RA. Does Prayer Influence the Success of in Vitro Fertilization–Embryo Transfer? Report of a Masked, Randomized Trial.
J. Reproductive Medicine. 2001; 46(9):781-87.
[17]
Aviles JM, Whelan E, Hernke DA, Williams BA, Kenny KE, O’Fallon M, Kopecky SL. Intercessory prayer and cardiovascular disease progression in a coronary care unit population: A randomized controlled trial.
Mayo Clinic Proceedings 2001; 76:1192-98.
[18]
Krucoff MW, Crater SW, Gallup D, Blankenship JC, Cuffe M, Guarneri M, Krieger RA, Kshettry VR, Morris K, Oz M, et al. Music, imagery, touch, and prayer as adjuncts to interventional cardiac care: The Monitoring and Actualisation of Noetic Trainings (MANTRA) II randomised study.
Lancet 2005; 36: 211–17.
[19]
Krucoff MW, Crater SW, Green CL, Maas AC, Seskevich JE, Lane JD, Loeffler KA, Morris K, Bashore TM, Koenig HG. Integrative noetic therapies as adjuncts to percutaneous intervention during unstable coronary syndromes: Monitoring and Actualization of Noetic Training (MANTRA) feasibility pilot.
Am Heart J. 2001; 142(5):760-69.
[20]
Benson H, Dusek JA, Sherwood JB, Lam P, Bethea CF, Carpenter W, Levitsky S, Hill PC, Clem DW, Jain MK, Drumel D, Kopecky SL, Mueller PS, Marek D, Rollins S, Hibberd PL. Study of the Therapeutic Effects of Intercessory Prayer (STEP) in cardiac bypass patients: A multicenter randomized trial of uncertainty and certainty of receiving intercessory prayer.
Am. Heart J. 2006; 151(4):934-42.
[21]
Crawford TC, Magruder JT, Grimm JC, Suarez-Pierre A, Sciortino CM, Mandal K, Zehr KJ, Conte JV, Higgins RS, Cameron DE, Whitman GJ. Complications After Cardiac Operations: All Are Not Created Equal.
Ann. Thorac. Surg. 2017; 103:32-40.
[22]
Astin JA, Stone J, Abrams DI, Moore DH, Couey P, Buscemi R, Targ E. The efficacy of distant healing for human immunodeficiency virus—Results of a randomized trial.
Alternative Therapies 2006; 12:36–42.
[23]
Rosa MI, Silva FR, Silva BR, Costa LC, Bergamo AM, Silva NC, Medeiros LRF, Battisti IDE, Azevedo R. A randomized clinical trial on the effects of remote intercessory prayer in the adverse outcomes of pregnancies.
Ciencia & Saùde Coletiva 2013; 18:2379–84.
[24]
Walker SR, Tonigan JS, Miller RW, Corner S, Kahlich L, et al. (1997). Intercessory prayer in the treatment of alcohol abuse independence: A pilot investigation.
Alternative Therapies in Health and Medicine 3(6):79-86.
[25]
Palmer RF, Katerndahl D, Morgan-Kidd J. A Randomized Trial of the Effects of
Remote Intercessory Prayer: Interactions with Personal Beliefs on Problem-Specific Outcomes and Functional Status.
J Alternative and Complementary Med 2004; 10:438–48.
[26]
Peter A. Boelens PA, Reeves RR, Replogle WH, Koenig HG. A randomized trial of the
effect of prayer on depression and anxiety.
Intl J Psychiatry in Med 2009; 39:
377–92.
[27]
Dehghani KH, Zare Rahimabadi A, Pourmovahed Z, Dehghani H, Zarezadeh A, Namjou Z.
The Effect of Prayer on Level of Anxiety in Mothers of Children with Cancer.
Iranian Journal of Pediatric Hematology Oncology 2012; 12:1–6.
[28]
Jadad AR, Moore AR, Carroll D, Jenkinson C, Reynolds JD, Gavaghan DJ, McQuayj HJ. Assesing the Quality of Reports of randomized Clinical Trials: Is Blinding Necessary?
Controlled Clinical Trials 1996; 17:1–12.
[29]
Simão TP, Caldeira S, Campos de Carvalho E. The Effect of Prayer on Patients’ Health: Systematic Literature Review.
Religions 2016; 7, 11.
[30]
Kaplan BH. A Note on Religious Beliefs and Coronary Heart Disease
J. South Carolina Medical Assoc. 1976, Feb. (suppl.), 60-64.
[31]
Graham TW, Kaplan BH, Cornoni-Huntley JC, James SA, Becker C, Hames CG, Heyden S. Frequency of church attendance and blood pressure elevation.
J. Behavioral Medicine 1978; 1:37-43.
[32]
Levin JS, Schiller PL. Is There a Religious Factor in Health?
Journal of Religion and Health 1987; 26(1):9-36.
[33]
Levin JS. Religion and Health: Is There an Association, Is It Valid, and Is It Causal?
Soc. Sci. Med. 1994; 38(11:1475-1482.
[34]
Hummer RA, Rogers RG, Nam CB, Ellison CG. Religious Involvemetn and U.S. Adult Mortality.
Demography 1999; 36(2):273-285.
[35]
Sloan RP, Bagiella E, VandeCreek L, Hover M, Casalone C, Jinpu Hirsch, T, Hasan Y, Kreger R, Poulos P. Should Physicians Prescribe Religious Activities?
N Engl J Med 2000; 342:1913-16.
[36]
Koenig HG, Georg LK, Hays JC, Larson DB, Cohen HJ, Blazer DG. The relationship between religious activities and blood pressure in older adults.
Int J Psychiatry Med. 1997; 28(2):189-213.
[37]
VanderWeele TJ, Yu J, Cozier YC, Wise L, Argentieri MA, Rosenberg L, Palmer JR, Shield AE. Attendance at Religious Services, Prayer, Religious Coping and Religious/Spiritual Identity as Predictors of All-Cause Mortality in the Black Women’s Health Study.
American J Epidemiol. 2017; 185(7):515-22.
[38]
Hardy SA, Nelson JM Moore JP, Ebstyne King P. Processes of Religious and Spiritual Influence in Adolescence: A Systematic Review of 30 Years of Research.
[39]
Strawbridge WJ, Shema SJ, Cohen RD, Kaplan GA. Religious Attendance Increases Survival by Improving and Maintaining Good Health Behaviors, Mental Health, and Social Relationships.
Ann. Behavioral Med. 2001; 23(1):68-74.
[40]
Beecher HK. The Powerful Placebo.
JAMA 1955; 159(17):1602-06.
[41]
Kienle GS, Kiene H. The Powerful Placebo Effect: Fact or Fiction?
J Clin Epidemiol 1997; 50(12):1311-18.
[42]
Burneo JG, Montori VM, Faught E. Magnitude of the placebo efect in randomized trials of antiepileptic agents.
Epilepsy Behav. 2002; 3(6):532-534.
A pooled placebo response of 12.5% (95% CI: 10.03-14.94%) was found in another meta-analysis:
Guekht AB, Korczyn AD, Bondareva IB, Gusev EI. Placebo responses in randomized trials of antiepileptic drugs.
Epilepsy Behav. 2010; 17(1):64-69.
[43]
Jørgensen KJ, Hróbjartsson A, Gotzsche PC. Divine intervention? A Cochrane review on intercessory prayer gone beyond science and reason.
J Negat Results Biomed 2009; 8:7.
© 2021 Daniel J. Castellano. All rights reserved. http://www.arcaneknowledge.org
| Home | Top |