From Randi on LKL (link above)
At a date and time convenient to you — and I know how very busy you are — we would randomly select one of those ten persons — by choosing a number from a hat.
Then, either you would call us or we would call you — your choice — you'd be given the gender, name, and age of the chosen subject, and you would do a reading over the telephone without getting feedback, that is, without doing questions-and-answers or asking for guesses to be accepted or refused. That reading could take a minute or two, or as much as half an hour — again, Sylvia, your choice, so that you could be sure that you've made "connection" with the subject.
When the reading is finished, you would so indicate, and the subject would then be asked to give a score to the reading, from zero to ten points.
Following that, we would contact, again in random order, each of the other nine persons for whom the reading was not done, and present them with either a transcript of the reading, or an audio tape of it, for them to also score from zero to ten.
Now, we should expect that the person for whom the reading was done would obtain a score, say, from six through ten, and — unless my "guessing game" scenario is correct — the other nine for whom the reading was not done, would have scores of zero to five.
But, to simplify all this, in order to beat 50-to-1 odds — which is much better than the thousand-to-one odds we usually require for such a test! — eight of those scores would have to be less than the score given by the person for whom the reading was actually done.
