With the advent of the computer, dice where largely replaced by electronic random event generators (REGs) for testing PK. These devices, which are often circuit boards that can be placed inside microcomputers, produce electronic noise that is determined randomly by the emission of an electron from a weak radioactive source. The nature of the noise at the moment of sampling is then converted by computer to (usually) a binary number - 1 or 0 - that is transformed into a display on the computer screen that tells the subject how well he or she is doing. The subject's task is to use PK to get the REG to produce more than 50% of the selected target number.
Radin and Roger Nelson (1989) published in Foundations of Physics a meta-analysis of 832 REG experimental series from 68 experimenters, conducted from 1959-1987. Of these 832 series, 597 were experimental (PK attempted) and 235 were control. Removal of the studies with extreme effects reduced the experimental sample from 597 to 496.
Although the reviewers did not report p-values, it is obvious from their graphs that the overall success rate was much grater than expected by chance for the experimental series, and very close to chance (nonsignificant) for the control series. The file-drawer ratio was 90 to 1: 90 chance studies needed for each published study to render the overall PK effect nonsignificant. There was no significant relationship between outcome and study quality, so the authors concluded that PK had been convincingly demonstrated using the more modern and reliable REG procedure.