tim (03/14/83)
A recent article on dowsing mentioned "the `psi-missing' ploy". This is something that was also mentioned in James Randi's excellent book "The Magic of Uri Geller" (in which he tells us how Geller does each of the things he does). Both sources mentioned it in a derogatory light. Psi-missing refers to results in psi experiments in which the results are significantly worse than those expected by pure chance. Misguessing all cards from a 50 card deck of 5 different cards would be an extreme example. Randi says that most people have a name for that -- "we call it losing". This is sheer emotionalism. If the results differ from chance significantly, then *something* is going on. I'm not trying to be an apologist. I know of no cases where psi-missing has been shown in a large enough number of trials to be significant. What I am trying to do is get people to approach things rationally. Tim Maroney
karn (03/15/83)
Gee, I would think that "psi-missing" meant that the person's "psi power" had gone AWOL that day.
henry@utzoo.UUCP (Henry Spencer) (07/23/83)
The problem with invoking the "psi-missing" phenomenon is that it is
just one of many ways to explain away a failed experiment. It is true
that if the score on your experiment is significantly worse than chance,
this is significant data. However, it is *still* a negative result on
*that* experiment. If any conceivable experimental result can be
explained away in a manner that doesn't conflict with the original hypo-
thesis, then what you are doing is not really an experiment at all.
The history of psi experimenting is rife with incidents of experiments
with superficially-negative results being explained away by looking for
some correlation, *any* correlation, in the data. If you look long
enough and hard enough, you will most assuredly find *something* that
looks non-random in any finite set of data.
There is a simple and well-known method of guarding against such
unscientific fudging, while nevertheless allowing for the difficulties
of not knowing exactly what you are looking for. Once you have run
your experimental trials, you split your data into two halves along
a dividing line that was fixed in advance. The second half of the
data is put to one side; you don't look at it just yet. You then
analyze the first half in any way you please, looking for any sort
of correlation. Having done this analysis and decided what sort of
correlation you are getting, you write down an exact statement of
the resulting hypothesis. Then you check to see if the second half
of the data confirms it. If it doesn't, your results are negative,
period -- no further shuffling of hypotheses or rearranging of the
calculations is allowed. The more careful psi experimenters use this
sort of method routinely, and their results to date are 100% negative.
--
Henry Spencer @ U of Toronto Zoology
{allegra,ihnp4,linus,decvax}!utzoo!henry