GodExamined.com

Eleven of the most-cited peer-reviewed studies that tried to measure whether one person's prayer changes another person's body — sorted into prayer given in the room (proximal) and prayer given from a distance the patient never knew about (remote / intercessory). Graded honestly: the positive results, the null results, and the two famous "miracles" that collapsed under fraud. This is what the data actually say — not what either a believer or a skeptic wishes they said.

Plain English

When scientists put prayer in a controlled trial, the result depends almost entirely on one variable: was the person praying in the room, or far away? The single study that objectively measured improvement — deaf and blind villagers in Mozambique whose hearing and eyesight were tested with instruments before and after hands-on prayer — was in the room. Every large, well-designed study of distant prayer for strangers — including the biggest one ever run, on 1,802 heart-surgery patients — found no benefit at all, and the two most spectacular "distant prayer works" headlines (a doubling of IVF pregnancies, a healing of AIDS patients) both collapsed in fraud and scandal. The honest reading is not "prayer is fake" and not "prayer is a proven medical device." It is this: the double-blind trial can measure a stranger praying by rote from another city — and finds nothing — but that is not the prayer the Bible actually describes. What science does robustly find is that prayer changes the person praying (see below). The reach beyond the room is real to the believer and invisible to the instrument — and the pattern in the chart above is exactly what you'd expect if the measurable part of prayer travels through presence, touch, and relationship.

The eleven studies at a glance

HistoricalVisual Every row is a real, locatable, peer-reviewed publication. "Grade" reflects design quality and how well the result survived scrutiny, not whether the result is flattering.

1. Two completely different questions get called "the prayer studies"

Study & year	Where	N	Design	Finding	Grade
Brown — STEPP 2010	IN ROOM proximal	24	Instrumented pre/post, hands-on prayer	Measured gains in hearing (p<0.003) & vision (p<0.02)	B★
Byrd 1988	Distant	393	RCT, double-blind, CCU	Lower composite "severity" score (p<0.01)	B
Harris 1999	Distant	990	RCT, double-blind, CCU	~10% better weighted CCU score (p=0.04); no LOS effect	B/C
Sicher / Targ 1998	Distant	40	RCT, double-blind, advanced AIDS	Reported fewer illnesses — endpoints changed post-hoc	C↓
Cha / Wirth / Lobo 2001	Distant	219	RCT, IVF, Korea	"Doubled" pregnancy (50% vs 26%) — fraud	F
Krucoff — MANTRA II 2005	Distant	748	RCT, multi-centre, cardiac	No effect of prayer on outcome	A
Benson — STEP 2006	Distant	1,802	RCT, 3-arm, multi-centre CABG	No benefit; "certain" group had more complications	A
Leibovici 2001	Distant	3,393	RCT, retroactive (prayer 4–10 yrs later)	"Significant" shorter stay — a deliberate paradox	⚠ probe
Astin / Harkness / Ernst 2000	Review	23 trials	Systematic review of distant healing	57% positive — "further study warranted"	B
Masters / Spielmans 2006	Review	14 studies	Meta-analysis of intercessory prayer	"No scientifically discernible effect"	A
Cochrane — Roberts 2009	Review	10 trials / 7,646	Cochrane systematic review	Equivocal; supports no recommendation either way	A

Logical Almost every argument about "whether prayer works" collapses two experiments that share a word but nothing else. Keeping them apart is the whole game.

In the room (proximal intercessory prayer). A person lays hands on or stands beside the sick person and prays. The patient sees them, hears them, is touched by them, and knows prayer is happening. This bundles the supernatural request with everything medicine already knows moves the body: touch, attention, expectation, hope, the relief of not being alone.

Not in the room (remote / distant intercessory prayer). Strangers in another building — often another country — are handed a first name and asked to pray. The patient never knows. This is engineered to strip out touch, expectation and placebo, so that only a non-physical, God-mediated effect could move the needle. It is the "clean" test of the supernatural claim — and it is also, as we'll see, a test of something the Bible never actually promised.

The reason the distinction matters: the only study in this entire literature that measured a hard, instrument-verified improvement was an in-room study. Every attempt to detect the pure not-in-the-room signal, once the study got big and careful, came back empty. That is either a fact about prayer, or a fact about what a double-blind trial can and cannot see. We'll argue it's mostly the second.

2. In the room — the one study that measured a change

Brown et al. — the STEPP study (Mozambique, 2010) IN THE ROOM

Southern Medical Journal 103(9):864–869 · Candy Gunther Brown, Indiana University · N=24

Researchers carried an Earscan-3 audiometer and standardised vision charts into rural Mozambique and tested 24 villagers reporting impaired hearing (14) or vision (11) before and after in-person, hands-on prayer by local ministers. Hearing thresholds and visual acuity were measured with instruments, not asked about.

Result: statistically significant improvement in auditory function (p<0.003) and visual acuity (p<0.02), with effect sizes the authors noted exceeded what suggestion/hypnosis studies typically produce.

Honest limits: only 24 people, no control group, no randomisation, no blinding, and a pre/post design that cannot separate a genuine change from arousal, expectation, or measurement conditions in the field. The authors say exactly this and call only for "further study." It is the strongest measured signal in the literature — and it is still a pilot.

Why this one is different. It used objective instruments on a hard sensory endpoint, not self-reported wellbeing or a soft complication score. And it was proximal — which is the only place in this whole field a measured effect ever showed up. Logical That is a clue about mechanism, not a proof of the supernatural: presence and touch are exactly the channels through which real, non-miraculous healing effects are already known to travel.

3. Not in the room — the early positives, and why they're contested

HistoricalRebuttal Chain Two famous trials reported that distant prayer helped heart patients. Both are real and double-blind; both have a soft underbelly.

Byrd (1988) — the study that started it all

Southern Medical Journal 81(7):826–829 · San Francisco General CCU · N=393 (192 prayer / 201 control)

Prospective, randomised, double-blind. Born-again Christians prayed daily, off-site, for patients they never met. The prayed-for group had a significantly better composite hospital-course "severity" score (p<0.01) — fewer needing antibiotics, ventilation, or diuretics.

The catch: the headline "severity score" was a post-hoc grouping of 26 outcome variables; only about 6 individually reached significance, and mortality, length of stay, and most endpoints did not differ. When you test 26 things, some cross p<0.05 by chance. Suggestive, not decisive.

Harris et al. (1999) — the replication attempt

Archives of Internal Medicine 159(19):2273–2278 · Mid America Heart Institute, Kansas City · N=990

Designed to replicate Byrd with a cleaner endpoint. The prayed-for group scored about 10% better on a weighted composite of CCU complications (p=0.04).

The catch: the trial's own primary, pre-specified measures — length of stay and an unweighted course score — showed no significant difference. The significant result came from a secondary weighted score. A partial, fragile hit.

Mathematical The multiple-comparisons problem in one line: run enough outcome variables and roughly 1 in 20 will clear p<0.05 with no real effect present. Byrd's 26 variables and Harris's multiple scores are exactly the setup where a "significant" finding is most likely to be noise. This is why the field kept building bigger, pre-registered, single-endpoint trials — and why those trials matter more than these.

4. Not in the room — the two "miracles" that collapsed

Rebuttal ChainFalsifiability The most viral "distant prayer works" headlines came from these two. An honest case has to be the first to tear them down — because a movement that keeps citing fraud discredits the true evidence around it.

Cha, Wirth & Lobo (2001) — the IVF "doubling" FRAUD

Journal of Reproductive Medicine 46(9):781–787 · Columbia University byline · N=219 Korean women

Claimed that prayer by Christians in the US and elsewhere nearly doubled IVF pregnancy rates for Korean women (50% vs 26%) who never knew they were in a study.

What actually happened: co-author Daniel Wirth was not a scientist — he later pleaded guilty to federal fraud (a $1M+ scheme) and went to prison. Lead-named author Rogerio Lobo removed his name, saying he'd only learned of it after the fact. The study ran with no informed consent. The Journal of Reproductive Medicine quietly pulled it from its website. It should never be cited as evidence for anything except how fraud enters a literature.

Sicher & Targ (1998) — the AIDS "distant healing" DISCREDITED

Western Journal of Medicine 169(6):356–363 · N=40 advanced-AIDS patients

Reported that distant healing by assorted "healers" produced fewer AIDS-defining illnesses and hospitalisations in a double-blind trial.

What actually happened: mid-study, the new HAART drug cocktail meant almost no one died — wrecking the original endpoint. Rather than report a null, the team (at Elisabeth Targ's urging) had a statistician mine the data after unblinding and swapped in a long list of secondary symptoms until something looked significant. That is textbook "HARKing" and p-hacking. The result is uninterpretable.

Brutal honesty: the two most impressive-sounding distant-prayer results in history are a convicted fraud and a data-mined artifact. Any case for God that leans on them is building on sand. Strip them out — as we do — and the remaining distant-prayer evidence gets weaker, not stronger. We report it anyway, because that is what honesty costs.

5. Not in the room — the big, rigorous nulls

Historical When the field finally built large, multi-centre, pre-specified trials, the effect vanished.

Krucoff et al. — MANTRA II (2005) NULL

The Lancet 366:211–217 · multi-centre, N=748 cardiac-intervention patients

The first large multi-centre trial of off-site intercessory prayer (Christian, Jewish, Buddhist and Sufi Muslim groups prayed). Prayer had no significant effect on clinical outcome. (A separate bedside music/imagery/touch arm — an in-the-room intervention — did affect distress.)

Benson et al. — STEP (2006): the biggest and best NULL / ADVERSE

American Heart Journal 151(4):934–942 · 6 centres, N=1,802 CABG patients

The largest prayer trial ever run, funded largely by the Templeton Foundation, with three arms: prayed-for-and-told-uncertain, not-prayed-for, and prayed-for-and-told-certain. Congregations prayed by name for 14 days.

Result: intercessory prayer made no difference to complication-free recovery. And the one group told with certainty they were being prayed for had more complications (59% vs 52%) — most plausibly performance anxiety ("am I so sick they called in the prayer teams?").

Logical Why STEP is the anchor. It is the study every side agrees was well built: huge, multi-site, pre-registered, one clear endpoint. It removes the multiple-comparisons loophole that inflated Byrd and Harris. And it found nothing — arguably worse than nothing. Any honest account of "the measured science of distant prayer" has to put STEP at the centre and let it set the tone.

6. What the whole body says at once — the reviews

Mathematical Single studies mislead; pooling many is the discipline. Three independent syntheses reach nearly the same place.

Astin, Harkness & Ernst (2000) — the cautiously-open one

Annals of Internal Medicine 132(11):903–910 · 23 randomised distant-healing trials

57% of trials (13 of 23) showed a positive effect; the authors judged the evidence promising enough to "warrant further study."

But this 2000 review predates STEP and MANTRA II and includes trials later discredited or superseded. Its optimism has not aged well.

Masters, Spielmans & Goodson (2006) NULL

Annals of Behavioral Medicine 32(1):21–26 · meta-analysis of 14 studies

Verdict, verbatim: "There is no scientifically discernible effect for intercessory prayer as assessed in controlled studies."

Cochrane review — Roberts, Ahmed & Davison (2009) EQUIVOCAL

Cochrane Database of Systematic Reviews CD000368 · 10 trials, 7,646 people

The gold standard of evidence synthesis: "the evidence does not support a recommendation either in favour or against the use of intercessory prayer." Some individual studies hint positive, most do not; they decline to call for more trials.

7. The retroactive-prayer bombshell — what the method can't see

Leibovici (2001) — prayer sent backwards in time

BMJ 323:1450–1451 (Christmas issue) · N=3,393 bloodstream-infection patients

In 2000, a physician took 3,393 patients whose blood infections had occurred in 1990–1996 — years earlier — randomised them, and had prayer said for one group now, for events already in the past. The prayed-for group showed a statistically significant shorter hospital stay and shorter fever back in the 1990s (p=0.01, p=0.04).

This was published as a deliberate reductio. Leibovici's point: either you believe prayer reached back through time to change a completed medical record — absurd — or you accept that a "significant p-value" in a trial of an effect with no plausible mechanism tells you the whole paradigm is being mis-applied.

Logical The double edge of Leibovici. To the skeptic it says: p<0.05 proves nothing when there's no mechanism, so ignore the prayer positives. But it cuts the other way too: the randomised controlled trial is built to test physical, repeatable, dose-response causes. It is the wrong instrument for a personal, relational, sovereign act that Scripture says is answered according to God's will, not the experimenter's protocol. Leibovici accidentally shows that the RCT can neither confirm nor refute prayer as the Bible frames it — it can only test a caricature of it.

8. What does replicate: prayer changes the one who prays

Cumulative Force The distant-stranger effect is invisible to instruments. But a different, larger, cleaner body of science — on the person praying — is robust and correlational, and it is where the honest empirical weight actually sits.

Documented, peer-reviewed, replicated (see the site's "Measured Efficacy" sources):

~33% lower all-cause mortality for frequent religious-service attenders over 16 years (Li, VanderWeele et al., JAMA Internal Medicine 2016; 74,534 women).
~7-year longer life expectancy at 20 for weekly-plus attenders (Hummer et al., Demography 1999).
~50% lower divorce risk with frequent attendance (Li, Kubzansky & VanderWeele, PLoS ONE 2018).
Lower depression, especially under high stress (Smith, McCullough & Poll meta-analysis, 2003; Miller et al. 2012).
Measurable brain changes from contemplative practice — cortical thickness, gray-matter, amygdala correlations (Lazar 2005; Hölzel 2010/2011). Note: these are secular meditation studies, not Christian prayer specifically.

The honest caveat, kept in view: every item above is correlational — healthier, more connected people may attend more, not only the reverse — and most measures religious practice/attendance rather than prayer alone. It shows prayer and worship are woven into a way of life that measurably flourishes. It does not prove a supernatural mechanism. Presented as more than that, it would be dishonest.

9. Why a null trial does not falsify biblical prayer

LogicalFalsifiability This is the section a skeptic will call special pleading and a believer will call obvious. Both deserve the argument laid out plainly, so it can be judged.

The experiment tests a claim the Bible never makes. STEP-style trials assume prayer is a force: fixed input (14 days, so many intercessors), fixed dose, mechanical output, indifferent to the heart of the one praying or the will of the One petitioned. Scripture describes prayer as the opposite — relational communion answered "according to His will" (1 John 5:14), sometimes yes, sometimes no (2 Cor 12:8–9), never on demand.

And it runs headfirst into an explicit prohibition. "You shall not put the Lord your God to the test" (Deut 6:16; Matt 4:7 — Jesus quoting it to refuse a staged miracle). A double-blind protocol designed to compel God to perform on a schedule for a p-value is precisely the demand for a sign that Jesus refused. A null result from such a design is theologically unsurprising — it may be the expected answer to a mis-posed question.

Held honestly, both ways. This argument correctly explains why distant-prayer RCTs come back null — but it also means prayer's efficacy toward others is not scientifically demonstrable, and we should stop claiming it is. That is the cost of taking the theology seriously: you forfeit the "science proves prayer heals strangers" talking point. What survives is defensible — measured in-room effects, robust benefits to the one praying, and a coherent reason the remote signal stays hidden — and it is stronger for being honest.

10. Cumulative verdict

Question	What the measured science says
Does in-room prayer measurably change the body?	One instrumented pilot (STEPP) says yes for hearing/vision — small, uncontrolled, unreplicated, plausibly via presence & expectation. weak-positive
Does distant prayer for strangers change hard outcomes?	No. The two biggest, best trials (STEP, MANTRA II) and two of three reviews say no; the positives were fragile or fraudulent. robust null
Does prayer change the person praying?	Yes, robustly but correlationally — longevity, marriage, mood, brain. Not proof of the supernatural. robust
Has science disproven biblical prayer?	No. It disproved prayer-as-a-vending-machine — a claim the Bible also rejects (Deut 6:16). out of scope

The measured science of prayer tells a consistent, unglamorous, honest story: the closer the pray-er, the more the instruments detect; the farther and more mechanical the test, the less — because the reach that matters most travels through presence and God's will, neither of which fits in a double-blind box.

Sources

Effect-on-the-pray-er figures (§8) are documented with full citations in the God Server's "Measured Efficacy" source set (evidence/sources/E_efficacy.json): Li et al. JAMA Intern Med 2016; Hummer et al. Demography 1999; Li et al. PLoS ONE 2018; Smith/McCullough/Poll Psychol Bull 2003; Lazar NeuroReport 2005; Hölzel et al. 2010/2011.

Does Prayer Change Other People? — The Measured Science, In the Room and Out of It