Randi proposed to test this scientifically. Four containers of water would be set up in a room; they would be in different parts of the room on separate tables. The practitioner would "charge" one of the jars, then leave the room. All four jars would be inside thin cardboard sleeves to prevent visual identification of the "charged" jar. The experimenters would randomly move the jars around each time, out of sight of the practitioners. Once the positions had been randomized, the practitioner would come back in and try to locate the "charged" jar using the dowsing rod. Simple and scientific.
Unfortunately, problems soon developed. It seemed that the test, even the participant's thinking about it, was causing all the water in the room to become charged. It would be impossible to distinguish the jars! The ad hoc excuses flowed like the water. The effect could not be measured in this way. All the water would become charged. And so on ...
NOTE: Here you see one property of pseudoscience. All failures must be explained, so any failure (or inability to perform) is immediately followed by an excuse (sometimes very creative).
Since we're talking about dowsing, we explored the concept. Dowsers claim to be able to find water (or many other things) by using a simple device like the dowsing rod we saw, a pendulum (string with weight), forked stick, or other prop. When the dowser locates the target, the device reacts and indicates a hit. At least that's the claim.
There's an interesting passage from pages 199-201 of Flim-Flam (by James Randi). It describes how Randi tested a dowser. The guy claimed to be able to find metal with his dowsing rods. The target was a pile of coins provided by the dowser himself. The test proceeded through several stages until it was pointed out to him that he had located the coins under a piece of paper while they were lying on a steel table which he had totally failed to detect.
Randi notes that this test was not ideal; it was done on a radio show and proper advance setup was not possible. Randi himself set up the test for the dowser, which required him to suppress any reactions while the test was in process. Any reaction could convey clues. The test was single-blind, which means that the test subject did not know where the coins were but Randi did. Ideally, a third party would set up each test so that neither the subject nor Randi knew. This slight weakness was described very clearly by Randi in the book.
Randi also tested two "psychics" in the Moscow area who had quite a reputation for being able to "read" an individual working only from a photograph. Randi deliberately avoided giving the readers any "feedback", or response, so as to avoid passing any information that way. The readers would have to get every last bit of information via psychic channels. They missed badly, getting nearly everything wrong. Many of the statements were quite general and tentative, as the readers were fishing for feedback. Not getting any, they were forced to guess. They missed the most important thing about the subject in the photograph.
This was an excellent example of "readers" who fish for information from the subject. In this case, they expected reactions and feedback from those who brought the photograph. This feedback would help with the reading. Of course it would! The key point is that, if they were TRULY psychic, they would get all the information they needed that way and it would ALL be correct.