Metratron:
I really want to keep this conversation constructive, so i hope you do not take this the wrong way. I do not have strong oppinions on Psi research like i do on eg. the buddhist wheel of pain (or whatever it is called) or the existense of the aboriginal dream-world.
When i wrote it would violate the laws of physics i meant just that - they would have to be augmented and our understanding of the brain would have to undergo a major paradigme shift. then it would be the new laws of physics ofcourse :-).
Regarding extraordinary claims bla bla bla - yah, i will give you it has become a bit of a coatrack on which to hang your brain. However, there is a degree of truth in it which i can explain if you want.
I will try to keep this conversation constructive by putting myself in 'the firing zone'.
Lets assume that YOU are a Psi researcher who wish to devise an experiment to discover Psi. More exactly, lets take the coin-example (ie. a psychic can predict coin flips, and in 10'000 flips the psychic will have 2 flips more right than chance dictate. That is a weak but detectable experiment).
Now you want to do your experiment RIGHT. I mean super, super right. The test subjects are isolated in bunkers 200m beneath the earth so they cant have radio contact with the surface and all is video monitored. the machine is callibrated by NSA an tested extensively, and the stream of random numbers is XOR'ed together with a stream supplied by Randi himself on a special sealed harddrive that can only be read once.
The experiment is guarded by heavy experimental protecols. Since psychic energy require much concentration, each 'session' with a psychic subject takes 4 hours, and you have made a prior agreement with the subjects that they get exactly 4 hours with the machine (corresponding to N coin flips). The 4 hours is important since that is the amount of time required to build up psychic energy.
At the end, you expect to detect the 2 extra flips with a p-value at about 1/1000.
All of this makes the experiment quite expensive to (staff, flying psychics in and out, video, equipment, etc), so its a bit of a one-shot thing. But if you keep to protecols, it will be definite proof one way or another - either Psi exist or it does not. Randi agrees that if you get the p-value he will give you the money (he, himself, test the experiment in many ways the months before the psychics come in).
..oo000oo...
3 months into the experiment, during the last 4-hour session of the day, a particular psychic comes in, lets call him Al-Kazah. He is very sure of himself, and claim he will "blow you away" and guess about 1 extra flip out of 10 on a good day, and today he feel charged. You put him in the bunker and tell him to commence.
As you watch him on camera you can see he does not seem to have a good day - he is far behind average. You expect about 2 subjects to do as bad as Al-Kazah during the experiments and dutifully record his abysmal record as the protecols dictate. Al-Kazah is not pleased, however. You can see him on camera as he become more and more agitated. He ask you if the machine is 'working' several times and 2 hours in he refuse to coorporate. He scream at you and tell you there is a 'daemon' in the machine and it has 'eaten' his powers - he accuse you of being in a pact with satan and you eventually have to remove him from the room and call in an ambulance. There are 2 hours worth of coin flips remaning in Al-Kazahs session and he will not be able to complete them.
This is a serious question: What do you do?
a) Trash Al-Kazahs data so far. Since the coin flips are statistically independent, you can scrap N flips today and it will in no way affect the flips you get tomorrow, the day after tomorrow or anything. You can then either extend the experiment with one more subject or, more likely, just calculate with one less subject overall, getting a slightly larger p-value, but you still expect it to be okay since you gave yourself some margin.
b) Trash the entire experiment, waving goodbye to the 3mil. the experiment has cost and the chance of winning the prize.
c) Call Randi and tell him what has happened. Propose to him that you take Al-Kazahs seat and finish the remaning 2 hours of his session. Since you are not psychic, you can argue your influence on the experiment is neglible - from Randis standpoint it will mean nothing at all if Al-Kazah or you finish the experiment, since neither of you have any powers (according to him). We assume you are certain Randi will agree to such a deal since he is a gentleman and want to laugh at you in the end, and you taking his seat will be recorded in the video and in your final report, just like good research dictate.
d) Something else. (describe)
Seriously, what approach would you take, assuming you are honest and want to take the approach that introduce no bias as far as detecting Psi goes? (others may answer as well, i will try to evaluate them from a statistical perspective later).