Don't get yourself all worked up over the double-blind stuff. The idea is simply not to allow him to get any clues to the locations by watching you. I'm quite sure that if you conduct a test in your back yard that you can watch him do this thing without tipping him off, especially if you stress to him that he's not to look at you or talk to you during the test. At the same time just be careful not to give any non-verbal cues such as staring at the targets or moving around. I think sitting in a lawn chair while wearing sunglasses is sufficient for such a test.
The actual test will depend on what he claims he can do. The most important thing for you to do is start with open tests. For example, get some foam cups. Put water in some and stones/sand in the others so they all cups are "heavy" enough. Place them around the room. Let him do his thing. I'm willing to bet he can find the water with no problem - his rods will twitch accordingly.
Now you've got a testable claim. To construct the test properly, understand that as a human, you are not random. Therefore, use something like a coin to determine whether a given location will have water or stones. That way you don't unconsciously follow a pattern.
You have to be concerned about the cups themselves offering clues (minor defects like pinch marks, warping), so don't reuse the same cups and lids from the open trial. Rather than use 20 cups, do five trials with four cups each (one in each corner for the room, for example). Do not give him feedback after each trial - wait until the end to tally the results.
When he's done, use these tables to judge his results. With 20 trials and 50% probability (water, no water), he would need to get 17 or more correct to beat 100:1 odds. He would need 19 or 20 to be 10,000:1 odds.
Of course, he'll get some other number. Use this binomial distribution calculator to figure out the odds. For example, suppose he got 13 correct. Plug 0.5 into the probability of success field, 20 into the number of trials and 13 into number of successes. Hit calculate. The very last field says that he had a 0.13 chance of getting 13 or more correct. Plug that number into Windows Calculator and hit the 1/x button. That will show you that he had a 1 in 7.6 chance by simply guessing, which is not far off from guessing one roll of a die.
BTW, in case it's not clear, the reason for doing the open test is two-fold. First, it gives you a testable claim. If he can do it without lids, then sure he can do it with lids, right? Second, it helps prove the point that his brain is making the rods move, not the water.
The actual test will depend on what he claims he can do. The most important thing for you to do is start with open tests. For example, get some foam cups. Put water in some and stones/sand in the others so they all cups are "heavy" enough. Place them around the room. Let him do his thing. I'm willing to bet he can find the water with no problem - his rods will twitch accordingly.
Now you've got a testable claim. To construct the test properly, understand that as a human, you are not random. Therefore, use something like a coin to determine whether a given location will have water or stones. That way you don't unconsciously follow a pattern.
You have to be concerned about the cups themselves offering clues (minor defects like pinch marks, warping), so don't reuse the same cups and lids from the open trial. Rather than use 20 cups, do five trials with four cups each (one in each corner for the room, for example). Do not give him feedback after each trial - wait until the end to tally the results.
When he's done, use these tables to judge his results. With 20 trials and 50% probability (water, no water), he would need to get 17 or more correct to beat 100:1 odds. He would need 19 or 20 to be 10,000:1 odds.
Of course, he'll get some other number. Use this binomial distribution calculator to figure out the odds. For example, suppose he got 13 correct. Plug 0.5 into the probability of success field, 20 into the number of trials and 13 into number of successes. Hit calculate. The very last field says that he had a 0.13 chance of getting 13 or more correct. Plug that number into Windows Calculator and hit the 1/x button. That will show you that he had a 1 in 7.6 chance by simply guessing, which is not far off from guessing one roll of a die.
BTW, in case it's not clear, the reason for doing the open test is two-fold. First, it gives you a testable claim. If he can do it without lids, then sure he can do it with lids, right? Second, it helps prove the point that his brain is making the rods move, not the water.
