I found this (and the preceding bit) noteworthy on two points; first in the obvious mathematical respect that explains the relationship between favored hypotheses and less favored hypotheses which are both supported by data;

Second, by the realization that researchers favoring ESP most likely fail to apprehend the hypothesis that they are testing, w/r to their critics. In the case in question, they collected 37100 predictions, which seems a little excessive considering it had essentially no persuasive power to skeptics.