Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site decwrl.UUCP
Path: utzoo!linus!decvax!decwrl!dec-rhea!dec-pbsvax!cooper
From: cooper@pbsvax.DEC (Topher Cooper HLO2-3/M08 DTN225-5819)
Newsgroups: net.philosophy
Subject: Negative deviations in psi experiments.
Message-ID: <1116@decwrl.UUCP>
Date: Mon, 28-Oct-85 14:02:07 EST
Article-I.D.: decwrl.1116
Posted: Mon Oct 28 14:02:07 1985
Date-Received: Thu, 31-Oct-85 22:15:44 EST
Sender: daemon@decwrl.UUCP
Organization: Digital Equipment Corporation
Lines: 115

>In any case, the whole basis of the research into such phenomena is as
>fundamentally flawed as religious belief (for many it seems to be a "modern"
>substitute for "ancient" religion, much as some people turn to eastern
>belief systems to substitute for the western ones).  Working from the
>assumption that the phenomena do exist (wishful thinking in and of itself),
>they engage in research to "prove" this, but along the way they interpret
>the data AS THOUGH THE CONCLUSION WAS TRUE, thus "proving" the conclusion.
>An example:  in ESP testing, some percentage of correct answers is considered
>statistically average (i.e., if you were to just pick randomly the likelihood
>would be that you would get, say, 20%).  A much higher percentage is taken
>as "evidence" of ESP, but then SO IS an extremely LOW percentage!  (Wow,
>that much of a deviation from probability?  It "must" be some psychic
>phenomenon involved!) [ROSEN]

The first sentence, besides being completely unproven, is completely
irrelevant.  As has been pointed out to Rich any number of times, the reason
that someone believes in a theory has nothing to do with the truth of that
theory.  I could, for example, make speculations as to the reasons for his
apparent fear of particular theories, but I won't, because those reasons are
not relevant to the validity of his arguments.

The second sentence is, in general, simply untrue.  Much current work in
the field starts with the "assumption" that psi exists (because those workers
feel that there is already sufficient evidence) and attempts to find out
something about its characteristics.  They do not, however, attempt to prove
the existence of psi based on that assumption.

This is an interesting criticism because it much more frequently applies to
the critics: i.e., they start with an explicit assumption of the impossibility
of psi phenomena and from that derive a proof of its non-existence.

The example given, even if valid, does not support the criticism made
previously.  It represents an accusation of poor statistical procedure,
rather than an example of using the conclusion as a premise.

In any case the example only serves to show the shallowness of the reading and
thought that Rich has devoted to this issue.  This same criticism has been made
over and over again by the critics.  It is answered, over and over again.
But the same criticism keeps coming up.  This is particularly frustrating since
the flaw in the criticism is so laughingly elementary.  Anyone who has taken an
elementary course in statistics should be able to spot it, certainly if they
make an attempt to read any of the experiments so criticized.

Imagine someone checking to see if a coin is biased and concluding that there
is strong statistical evidence that it is.  Then a critic comes along and says
"A much higher than expected number of heads is taken as evidence of a biased
coin, but then SO IS an extremely LOW percentage!",  with the implication
that this invalidates the test for bias of the coin.  Of course it does not.
What is called a "two-tailed" statistical test is used.  A one-tailed test
would ONLY check for a higher than expected number of heads, or ONLY check
for a lower than expected number of heads (= higher than expected number of
tails).  A two-tailed test checks for both.  The cost of checking for both is
that you need a larger deviation in a particular direction to conclude bias,
than you would for a one-tailed test for that direction.  All this is standard,
elementary hypothesis testing theory from statistics.

In the early days of parapsychology, one-tailed tests were used exclusively.
The results were very inconsistent.  Then it was noticed that some subjects
in some conditions seemed to be getting fewer than expected hits, and that
these tended to cancel out the successes of other subjects in other conditions.
Without worrying about why this took place, the solution was to use two-tailed
tests (including intrinsically two-tailed tests such as analysis of variance)
ALL THE TIME.  This is now the policy, two-tailed tests are usually used in
parapsychology unless there is a particular reason not to.  Results are still
inconsistent (this problem not, apparently, being the only one) but much less
so.

"But," critics frequently respond, "how do we know that the parapsychologists
don't first try a one-tailed test for psi-hitting, and only when that fails
apply a two-tailed test?"  One answer is to simply look at the experimental
literature.  One frequently finds experiments reported with a "higher than
expected number of hits" which report the results in terms of probabilities
derived from two-tailed tests, resulting in lower overall statistical
significance than if a one-tailed test had been used.  Indeed, one rarely
sees one-tailed tests used at all.

What is the source of psi-missing (which is the term used to describe the
situation when a subject guesses significantly fewer targets than can be
ascribed to chance)?  If I could give a complete answer to that I would be
one of the top parapsychologists in the world.  A number of circumstances
under which psi-missing occurs have been found, however.  I will describe two
important ones.

Before an experiment, the subjects are asked "Do you think that psi exists?"
(in practice many variants of this question have been explored).  Some answer
"Yes, definitely", some answer "I don't know" and some answer "Definitely not".
Those who give one of the first two answers are labeled "Sheep", those who
give the last answer are labeled "Goats".  When the tests are given and
evaluated, it is frequently found that the Sheep tend to psi-hit, while the
Goats tend to psi-miss.  This is called the Sheep-Goat effect.

When an experiment is performed and there are two conditions used (i.e., test
and control conditions), and the subjects are aware of the distinction between
the conditions (e.g., one condition is "lights are on" and the other is "lights
are out" in the room the subject is in), then it is frequently found that one
condition will produce psi-hitting and the other will produce psi-missing.
There is some indication that the psi-hitting generally occurs in the condition
that the particular subject prefers.  This is one of several related effects
called in parapsychology, the "Differential effect."  For those who remember
the posting which started this (I won't bother to quote it) it should be clear
how this effect might call into question the meaningfulness of Hawkings
observation.

A number of other conditions have been found which seem to relate to
psi-missing, including some correlations with mood or personality scales, but
no-one would claim to be able to reliably predict when it will occur.  Lacking
such reliable predictions, two-tailed tests are used and rightly so.

		Topher Cooper

USENET: ...{allegra,decvax,ihnp4,ucbvax}!decwrl!dec-rhea!dec-pbsvax!cooper
ARPA/CSNET: cooper%pbsvax.DEC@decwrl

Disclaimer:  This contains my own opinions, and I am solely responsible for
them.