Several substantial studies have examined whether prayer for other people affects their medical outcomes. The results have been used by both believers and sceptics, and the studies themselves are more interesting than either use.
The design
The general approach: recruit patients undergoing a procedure, randomly assign them to be prayed for or not by groups of volunteers, and compare outcomes.
Some studies also varied whether patients knew they were being prayed for, which turned out to matter.
The findings
The largest and most rigorous studies have generally found no effect of intercessory prayer on clinical outcomes.
One substantial trial produced a result that generated considerable discussion: patients who knew they were being prayed for had somewhat worse outcomes than those who didn't know.
The proposed explanation was psychological — knowing you'd been selected for prayer may have suggested your condition was serious, producing anxiety. Whether that explains it is uncertain, and the finding is a useful reminder that expectation affects outcomes in ways that cut in unexpected directions.
Earlier and smaller studies reported positive findings, and these have generally not replicated in larger, better-controlled work, which is a familiar pattern across medical research.
The methodological problems
Substantial enough that the whole enterprise is questionable, and worth spelling out.
Contamination. Patients in the control group were almost certainly being prayed for by family and friends. There is no unprayed-for control group, which means the study compares more prayer against less prayer rather than prayer against none.
Dose. Nobody knows what quantity or quality of prayer would be expected to produce an effect, so the intervention cannot be specified.
Mechanism. Without a proposed mechanism, there's no basis for predicting what outcomes to measure, over what period.
The theological objection. Many religious traditions would reject the premise. Prayer is generally not understood as a mechanism producing predictable effects, and several theological positions hold that testing in this manner is inappropriate.
Which means a null result doesn't refute a claim most believers were making, and a positive result would have been difficult to interpret theologically.
What is better established
Research on personal religious practice and health outcomes is a different and larger literature, with more consistent findings.
Associations have been reported between religious participation and various health measures, including mortality, in a number of large observational studies.
Interpretation is where the difficulty lies. Proposed explanations include social support from religious communities, behavioural differences including lower rates of substance use, coping resources, and a sense of meaning.
Confounding is substantial — people who attend religious services differ in numerous ways — and causal claims are correspondingly weak.
What can be said reasonably: participation in a community with shared practices is associated with better outcomes, and much of that may be attributable to the community rather than to anything specific to religion.
Prayer as a practice
Setting aside outcome studies, the question of what prayer does for the person praying has a different character.
Contemplative prayer, in the traditions that develop it, has substantial overlap with practices studied under other names. Repetition of a phrase, sustained attention, deliberate cultivation of particular states.
Research on these practices generally finds effects on stress and mood, with the caveats about methodology that apply throughout this area.
Petitionary prayer — asking for something — has been examined less. Some work suggests it functions partly as a form of rehearsal and articulation, in which the person clarifies what they actually want, which is not nothing.
And prayer as expression of gratitude connects to a literature on gratitude practices, which has produced reasonably consistent findings on wellbeing measures, with the usual limitations.
The category error
The underlying issue with the outcome studies, and it's instructive beyond this topic.
They treat prayer as a causal intervention, testable like a drug. That framing suits neither the practitioners, who mostly don't understand it that way, nor the sceptics, for whom no result would be persuasive.
Which is why the studies satisfied nobody. They answered a question that was constructed for the purposes of testing rather than one anybody was actually asking.
The more interesting questions — what does the practice do for the person, what does communal observance do for a community, how does a framework of meaning affect how people cope — are harder to study and considerably more informative.
Where that leaves it
Intercessory prayer has not been shown to affect clinical outcomes, in studies with methodological problems severe enough that the result may not mean much.
Religious participation is associated with health outcomes, probably substantially through social and behavioural pathways.
And the practice itself, for those who undertake it, is doing something the outcome studies never attempted to measure — which is arguably the appropriate conclusion for a question that was never well suited to a randomised trial.
Prayer in illness, from the patient's side
A different research question, and one with more consistent findings. Studies examining what patients themselves report have found that a substantial proportion pray during serious illness, including many who do not otherwise describe themselves as religious.
What they describe it doing is generally not what the outcome studies tested. Accounts cluster around coping, around articulating fear, around a sense of not being alone with it, and around maintaining hope without denying the situation.
Whether those functions could be served another way is a fair question. What is clear is that they are different functions from the causal one that trials attempted to measure, and that patients were mostly not making the claim being tested.