The Science of Prayer
Praying for yourself may help you. Or it may harm you. Or it may do nothing at all. Each of these is possible, by purely material brain-body interactions with nothing supernatural required. And, praying for another person with his or her knowledge might also help, again by purely material means, in reducing that person’s stress. But it can also hurt by adding more stress. Many studies on prayer and health can be found in the literature and I do not have the space to review them all. Instead I will focus on the popular book Healing Words by physician Larry Dossey in which he reports on “an enormous body of evidence: over one hundred experiments exhibiting the criteria of good science, many conducted under stringent laboratory conditions, over half of which showed that prayer brings about significant changes in a variety of living beings."1 One wonders why he would even count those that were not conducted under stringent laboratory conditions.
Dossey refers to a survey by Daniel J. Benor, M.D., published in the journal Complementary Healing Research, of experiments dealing with healing effects of prayer on enzymes, cells, yeasts, bacteria, plants, animals, and human beings.2 According to Dossey’s summary of Benor’s results, researchers have performed 131 controlled trials. Of these, fifty-six show “statistically significant results at a probability level of <0.01 or better (that is, the likelihood that the results were due to chance was less than 1 in 100).” another twenty-one studies “demonstrate results at a probability level of 0.02 to 0.05 or better (that is, the likelihood that the results were due to chance was between 2 and 5 chances in 100).”
Dossey is incorrect in his interpretation of the statistical significance of these experimental results, making a common mistake one finds in many papers and books. The “probability level” quoted in most scientific papers is usually what statisticians call the “p-value.” For example, suppose an effect is reported with a p-value of five percent. This means that in a long sequence of identical experiments we would expect to observe an effect as great or greater produced by statistical fluctuations in five percent of the cases. This not the same as “the likelihood [or probability] that the results were due to chance.” In fact, it is always possible to get any observed effect by chance. You simply have to repeat the experiment enough times.
In any case, Dossey informs us that “ten of the studies are unpublished doctoral dissertations, two are masters’ theses, and the rest are published primarily in parapsychological journals.” He asserts that “these publications have peer review standards as rigorous as many medical journals.”
However, as I have mentioned before in this column, the standards of medical journals are quite low compared to other science fields such as physics. This is presumably necessary to assure that useful therapies are not kept from needy patients for too long. Unlike physicians, however, physicists and parapsychologists are not in the business of saving lives but rather that of investigating extraordinary phenomena. Those who search for evidence of psychic or spiritual phenomena should be bound by the stricter standards of physics and other fields which deal with extraordinary claims. After all, the scientific confirmation of such phenomena would be of world-shaking significance.
No respectable physics journal would publish a result with a p-value of one percent. If it did, every hundredth paper or so would contain a false claim that was only a statistical artifact, wreaking havoc with the whole research enterprise. In fact, the publication standard in physics is typically a p-value of 0.01 percent, that is, only one in 10,000 similar experiments would be expected to produce the reported effect or a greater one as a statistical fluctuation. If this standard were applied to Dossey’s sample, none of the 131 trials mentioned above would be published.
The same can be said for all the intercessory prayer studies that have been published in medical journals, accompanied by great media hype. For example, cardiologist Randolph Byrd has claimed evidence that coronary patients benefitted from blind, distant intercessory prayer. But his p-value is only five percent.3 Such results would be expected from statistical fluctuations alone every twenty experiments, on average. Another study along the same line as Byrd’s has been published in a major medical journal, Archives of Internal Medicine, with nine coauthors.4 There, positive results are reported at a p-value is four percent, but for different criteria than Byrd’s. In fact, they fail to confirm Byrd’s specific results.
Dossey is simply wrong when he says the evidence is “simply overwhelming that prayer functions at a distance to change physical processes in a variety of organisms, from bacteria to humans.” Even without examining the detailed protocols of these experiments, the statistical significance is insufficient to draw such a conclusion. We have no idea how many experiments may have been done that gave no positive effects and consequently were never published (the “filedrawer effect”). These papers should not have been published either.
This article is abstracted from my latest book, Has Science Found God? The Latest Results in the Search for Purpose in the Universe, to be published by Prometheus Books. Thanks to Bill Jefferys for helping me clarify the statistical issues.
- Larry Dossey, Healing Words: The Power of Prayer and the Practice of Medicine (San Francisco: Harper, 1993).
- Daniel J. Benor, Survey of spiritual healing research. Complementary Medical Research 4, no. 1 (1990):9-33.
- Randolph C. Byrd, Positive therapeutic effects of intercessory prayer in a coronary care unit population, Southern Medical Journal 81, no. 7 (1988):826-29.
- W.S. Harris, M. Gowda, J.W. Kolb, C.P. Strychacz, J.L. Vacek, P.G. Jones, A. Forker, J.H. O'Keefe, and B.D. McCallister, A randomized, controlled trial of the effects of remote, intercessory prayer on outcomes in patients admitted to the coronary care unit, Archives of Internal Medicine 159 (1999): 2273-8.