by Tom Napier,   Copyright © 1998  This page posted at
http://www.voicenet.com/eric/z/telepath.htm

There are strong physical grounds to doubt the existence of telepathy.
However, if it exists, how can we demonstrate it?  What would be
acceptable evidence for telepathy?  Is there a fool-proof device to test
telepathic ability?

Were telepathy strong and reliable it would be easy to recognize it and
there would be no doubt about its existence.  The case for telepathy is
that it is fleeting and unreliable and we do not yet understand it well
enough to be able to specify under what circumstances it will occur.  Thus
the best we can do is to have two people attempt to communicate and to
look for deviations between their actual results and those to be expected
by guessing.  Unfortunately, some subjects will be motivated to exploit
any flaws in the experimental protocol which allow them to mimic
telepathic communication.  The experimenter's job is to eliminate all such
flaws.

 Here comes Catch-22.  Define telepathy as mind-to-mind communication
without a physical link.  Proving telepathy exists requires, 1) Evidence
of communication and 2) Evidence that there is no conceivable conventional
method of communication.  Proving the former ought to be simple, though it
should be noted that many telepathy tests have foundered on this reef.
Proving the latter is extremely difficult, perhaps impossible.  Ergo, the
existence of telepathy can never be proved to a skeptic.  Undeterred, let
us advance upon the impossible.
 
 

Selecting the test information
 
 

The object is to confirm that information given to one person can be
recovered from another person who is not in communication with the first
by any normal means.  This requires that the second person have no means
of knowing the test information in advance and also that there be no means
of deducing some part of the information from the rest.  For example, one
would not expect to prove telepathy by having one person transmit, and the
other receive, a list of the presidents of the United States.  The test
information should be random and ideally it should not exist prior to the
test.

Since telepathy, if it exists, is a weak and unreliable method of
communication it is difficult to distinguish it from "just guessing."
Thus we need to use a test protocol which would give well defined results
in the absence of telepathy so that we can recognize the presence of
telepathy by its significantly different test result.  The test data must
thus be statistically well behaved.

It must also be possible to make an unambiguous comparison between the
transmitted and received information.  This rules out any experiment, such
as the "ganzfeld" type of experiment in which the accuracy of a "hit" is a
matter of subjective interpretation.  It is difficult to tell if results
are "better than chance" when one cannot compute the expected probability
in the absence of telepathy.  The old zener card experiments, for all
their practical flaws, were superior in the methodological sense and will
be used as a model for the present purpose.

(Since card-guessing gives test results which lend themselves to
accurate statistical analysis, whereas ganzfeld experiments do not, the
present fashion for ganzfeld experiments is rather puzzling.  It suggests
that investigators prefer fuzzy experiments which give positive results to
well-designed experiments which give null results, a trend counter to that
universally adopted in real science.)
 
 

Assessing the results
 
 

In telepathy tests even the best results differ little from what would
be expected from random guessing.  High scores are rare, consistent high
scores are almost unknown.  For a deviation from chance to be significant
we need to carry out many tests Unfortunately, no score in a telepathy
test "proves" telepathy. If you guess a pack of 25 zener cards randomly
(For example, not guessing each type of card a total of five times.) then
at one test a minute you will guess all 25 correctly in around 20 times
the age of the universe.  What if you get all 25 right first time?  That
could be just "chance."  Do it again.

The classical tests used a freshly shuffled pack of cards which could be
assumed to be in a random order.  Both the sender and the receiver had to
make a list of the cards so that the lists could be compared after the
experiment.  The distribution of "hits" to be expected in the absence of a
cause can be computed and the probability of any excess number of hits can
be determined.  This, of course, only indicates the probability of that
result occurring in the absence of telepathy.  It should not be
interpreted, as was done by early experimenters, as the probability that
telepathy does not exist.
 
 

Practical experiments

 I first studied telepathy testing about the middle of the long period
between Soal's experiments and the emergence of definitive proof that he
had cheated by altering the target lists after the experiment.  Books on
telepathy referred to Soal's tests, with their smaller and smaller
probabilities of the results being "due to chance," as better and better
evidence for telepathy.  The authors were right in a sense.  There was a
cause behind the anomalous results; they only erred in thinking that the
cause was telepathy.

While I did some card-guessing experiments when I was a physics student
I did not seriously address the problem of designing a telepathy
experiment until the mid-1970s when an acquaintance, who believed he had
telepathic powers, offered himself for testing.  I designed a tester and
built it from parts in my junk box.  After several test sessions we
concluded that my acquaintance had failed to demonstrate anomalous
communication.

The design considerations I addressed then are just as valid today.
Rather than transmit one of five card patterns I elected to use one of
four colors as targets.  (This requires that at least the sender have
normal color vision.)

The tester had two parts, a sender's box and a receiver's box. They
were linked by a 25 foot cable.  This was (just) long enough to permit the
sender and receiver to sit in separate rooms with the communicating door
closed.  A serious tester would use a much longer cable and better
acoustical isolation between the subjects.

A panel in the sender's box was illuminated half the time by one of
four colored bulbs, one red, one yellow, one green and one blue.  The
other half of the time the panel was dark.  The rate at which the panel
cycles between light and dark could be adjusted to suit the sender.

The receiver's box carried four push-buttons with four colored tops.
These matched the colors of the four bulbs in the sender's unit.  At the
moment when the sender's panel was illuminated, the receiver's box beeped
to signal to the receiver that the sender is looking at a colored panel.
The telepathic receiver then tried to push the corresponding colored
button.

If the correct button was pressed a "hit" was recorded on an
electromagnetic counter.  If any of the three incorrect buttons was
pressed a "miss" was recorded.  If the receiver failed to press a button
nothing happened.  The test was stopped after a predetermined number of
trials and the number of hits and misses were noted.

Since there are four possible colors, one would expect, in the long
term and in the absence of telepathic ability, that the number of hits
would be about a third the number of misses.  Any large deviation from
this ratio indicates either that an unusual but random event has occurred,
or that the sender and receiver have been able to communicate in some
manner.
 
 

Design considerations
 
 

This design took some pains to avoid the pitfalls into which many
experimenters have fallen.  For example, there was no list of targets
which might not be random, which might have been seen in advance by the
receiver or which might be altered after the experiment.  The target
colors were chosen by a mechanism which was completely random and which
was equally likely to pick any of the four colors.  (One research
organization's tester rejected repeated targets, skewing their results.)
There is no means by which the next color to appear can be predicted
before it is presented to the sender.

There is no list of the receiver's choices which might be recorded
incorrectly or be altered by the person checking the results.  The
receiver's choice is compared immediately and automatically with the
selected color and is recorded by the counters as a hit or a miss.

The timing of the trials is controlled electronically.  This eliminates
the possibility of the sender and the receiver communicating by a code in
which the time at which each trial is announced conveys information.
Synchronization between the sender and the receiver is achieved by an
electronic beep.  No human agent need communicate between the two parties
and thus possibly give clues to the correct color.

A button is only recorded when a color is being displayed.  Pushing
more than one button or holding down a button does not result in a guess
being registered. (I had to modify the original design when I found that
pressing all four buttons at once always registered a hit. Now only the
first press in any cycle is counted, later presses are ignored.)

This device has four known flaws.  1) The cable between the units is
not long enough to make it practical to maintain total acoustic isolation
between the sender and the receiver.  For this reason I would not treat a
positive score as being strong evidence for telepathy.  2) A test run
should consist of a preselected number of trials; this box indicates only
hits and misses and thus requires a referee to watch the counters and to
stop the run when their total reaches the chosen number.  3) The counters
are resettable; this might permit a third person to reset the miss counter
unobserved to achieve a high ratio of hits to misses.  4) Though the
counters are not visible to the sender, they make a distinctly different
noise when recording a hit and when recording a miss; this permits the
sender, at least, to know whether the receiver's choice was correct.
Since the sender cannot influence the choice of targets this should not
affect the result but it would have to be eliminated in a more serious
tester.  An updated design, not yet built, displays only the total number
of tests, stops automatically after a predetermined number, and then
indicates the number of hits.

It might also be noted that we now live in an age when small electronic
transmitters and receivers are readily available.  Test subjects should be
thoroughly searched before each experiment.
 
 

Randomizing the targets
 
 

To select one of a number of events completely at random is an
interesting technical problem which not all experimenters have succeeded
in solving.  It cannot be done with a computer, their "random" numbers are
completely predestined.  It would always be possible, given a sequence of
their random numbers, to predict all possible future random numbers.  The
best answer is to use a "noise diode" which generates a signal which
depends on the random motion of electrons.  The output of such a diode is
always time-correlated to some degree, it can't change infinitely quickly,
but it can, for example, generate very rapid pulses.  The number of pulses
occurring in a long time interval can be treated as a random variable with
a wide range around a mean number.  If this number of pulses is divided by
a small number, four in this case, the remainder should be uniformly
distributed.  This is the technique used in the telepathy tester.  It
generates and counts about a million pulses a second.  If the two-bit
counter is stopped every few seconds the number it contains is equally
likely to be 0, 1, 2 or 3.
 
 

Chance results
 
 

With four randomly selected colors the expected chance result is, of
course, 25% hits.  Because the colors are selected completely at random,
unlike in the zener card experiments in which each symbol appears exactly
five times in each test, it is particularly easy to use the binomial
theorem to predict the distribution of the results.

Since the object is to demonstrate communication, only hit numbers
greater than 25% are significant.  Modern researchers should not fall into
Rhine's "psi-missing" trap and start seeing significance in lower than
expected scores.  The table below shows the probability of getting N or
more hits in a run of 20 trials.  5 hits is the expected number.
 
 
 
 

    Hits     Probability              Hits     Probability
 
 

     6     3.828273e-01               14     2.951175e-05

     7     2.142181e-01               15     3.813027e-06

     8     1.018119e-01               16     3.865316e-07

     9     4.092517e-02               17     2.960496e-08

    10     1.386442e-02               18     1.610715e-09

    11     3.942142e-03               19     5.547918e-11

    12     9.353916e-04               20     9.094947e-13

    13     1.837041e-04
 

Conclusion
 
 

With a little care a telepathy tester can be built which is immune to
the more common experimental flaws.  Unfortunately it may succeed only in
demonstrating the absence of communication, not the presence of telepathy.
In this case, experimenters should avoid the temptation to redefine the
"correct" protocol to be that one which generates positive results.
Negative results are also significant.
 

 More rants by Eric Krieg and Tom Napier  Eric Kriegs page on how to test Telepathy