The Effect Was in the Method

Prayer for the sick makes contact with something measurable. Somebody either recovers or does not. Hospitals keep records. So the claim can be run as a trial, and it has been, more than once, by researchers who expected it to work.

Taken one study at a time the results look inconclusive. Taken in order they do not.

A real effect behaves in a known way as the method improves. It survives blinding. It holds when the patients do not know, when the staff do not know, and when the outcome is fixed in advance rather than chosen once the data is in. Tightening a procedure strips out noise. That is the whole reason for tightening it.

The prayer literature runs the other way. The early positive findings came from small studies with outcome scores devised for the occasion. The larger and better controlled the design, the nearer the result sat to nothing. A trial of cardiac bypass patients reported in 2006, eighteen hundred of them across three arms, found no difference in complications between the group prayed for and the group not. The one arm that fared slightly worse was the arm told with certainty that prayer was being offered for it.

The standard reply is that the design insults the premise. A deity is not a reagent and does not perform on demand, and petitions read off a card by strangers are not what the tradition means by the word. The objection has force. It also arrived after the results did. The same trials were welcomed while they were running, and the early positive findings were cited widely with no such caveat attached.

Notice what the reply costs. Withdrawing the claim from measurement protects it from disconfirmation and cancels its confirmation in the same motion. A prayer that cannot fail a trial cannot pass one. Recoveries credited to it afterwards are being counted by a method the position has just ruled inadmissible.

None of this says nobody should pray. Prayer does other work, some of it measurable too, and people who pray report less distress under strain. That is a finding about the person praying and needs nothing outside the room to explain it. What the trials rule out is narrower. Intercession at a distance, offered for strangers, does not move surgical outcomes at any rate anyone has managed to detect.

There is a test in this and it needs no hospital. Find a claim about prayer that names its outcome before the outcome is known. Then check whether the naming survived contact with the result, or whether it was quietly adjusted once the result was in. A claim that only ever reads backwards has not made a prediction.