Here you can find the reference, Amos. This is a review article and the original studies are referencend in it.

The good news is that the placebo effect can make us feel better, and the bad news is that in the worst case that may be all it can.

Again with the example of the one selling a petrol additive for more mileage:
- We know for sure that a positive effect (of the additive) is there if we don't look at the actual miles done with the car but ask the driver how it feels driving with the additive (and the driver says, yes it feels better, whether the additive is in there or not)
- We suspect that there may be no effect upon the mileage at all and that everything the buyer gets for the money is a better feeling.
- There could be a (much smaller than placebo) real effect lurking. Perhaps a driver with the additive thinks more about saving gas and therefore drive a little different which would make a difference even with an objective outcome measure. But we would know then that not the additive but the behaviour change is the relevant thing.

What always surprises me in these discussions (about nonconventional medical treatments) is the following: The believers say it works and when asked how they know they say it's their experience. When asked what are the constituents of their experience they say it comes down to counting cases where it has worked and compare it with cases where it has not. Basically, they do a (informal, more like estimating, but humans are not bad at estimating numbers) counting procedure and nothing else.

Scientists do more or less the same with some minor differences:

(1) They replace estimating by actually counting
(2) They try to define before counting what constitutes a success (or a failure)
(3) They try to control some biases in judgement by proper design (for instance, they do not let someone with a bias do they counting or at least let the one doing the counting not know what she counts, so any bias doesn't interfere with the results)

That's basically how these studies are done: Both ask the same question, namely, does it help (on the average) and both let the same procedure, namely (implicit or explicit) counting, decide.

The interesting difference comes when the one procedure ('experience') leads to positive results and the other (scientific study) leads to a result of no effect. Then the believers say, your procedure doesn't grasp the realities of life and all that, and forget that the scientists have done more or less the same what they do. The scientists point to the possibility that the differences (better control, no influence of bias,...) may account for the results.

The believers often cannot even admit that possibility for they have invested too much (financially and emotially) and perhaps profit a lot (think again of the gas addtitive salesman; would he be interested in a study showing it doesn't help the buyers, only the sellers?). They'll use all kinds of ad hoc explanations to avoid the one they fear most.

Wolfgang