In the initial PK experiments with pre-recorded targets, great care was taken that no human observer had seen the pre-recorded events before the subject had made the PK effort (JASPER, 1976 267-291). In a later study, however, where the initial observer saw all details of the pre-recorded random sequences and the final scores, the observation led to a significant reduction of the subject's PK success

(RIP 1984, 22-23). Particularly interesting was the fact that the observers had no apparent PK abilities.

This led to the question of whether nonhuman first observers also might produce such a PK-inhibition effect. A pilot study in this direction used a dog as observer (RIP 1983, 96-98). The following study used goldfish subject to weak electric shocks.

The basic test arrangement was very simple. We first prerecorded random events (heads and tails) for many test runs. Next we displayed half of the test runs to goldfish. For this purpose, we let a computer slowly read the pre-recorded sequence and administer for each head a weak electric shock to the fish. Finally, we displayed all test runs to human PK subjects while they were trying to enforce the appearance of many heads. The PK subjects were kept unaware of whether or not a run had been previously displayed to the fish.

To evaluate the results, we checked whether the human PK subjects received higher scores on those runs that had not been displayed to the fish than on runs that had been displayed.

The whole study consisted of two similar parts, using as PK subjects a martial arts teacher (Mr. Guy Savelli) and myself, respectively. Each of these parts was structured into eight blocks, or miniexperiments, of 128 test runs each, with each test run comprising 128 individual binary decisions.

At the beginning of each block, I used a Geiger counter connected to a small computer to generate the required 16,384 random bits. As a safeguard against malfunctions of the random generator, I used random number tables to randomly invert (interchange heads and tails) sections of the initial sequence. I then stored the resulting bit sequence in a permanent memory chip (EPROM 2532).

A total of eight 256-bit sections were presented to the fish in one session. Each section took about two minutes. The sections were separated by 20-minute intermissions. Two groups of four goldfish each were used in alternation. Each group never had more than one session within a 24-hour period, and, between blocks, the fish usually had breaks of many days or weeks.

The fish were common Comet goldfish with a body length (excluding the transparent tail fin) of about five centimeters. They were tested in their home aquariums, after copper plates had been inserted at the ends. The aquariums were 16" and 8" wide, with a water level of 7". The inserted copper plates covered the whole end areas. The shock was provided by a 60 Hz pulse of 1/10 second duration. The voltage was set between 3 and 9 volts such that the fish displayed only a slight reaction to the shocks. The peak current at 6 volts was 0.16 amps. The shocks were disturbing to the fish to the extent that they would stop feeding during the two-minute test intervals, but there was no apparent after-effect.

After half of the runs had been thus displayed to the fish, the PK test runs with the human subjects could begin. The basic PK goal was always the same: to observe in the pre-recorded sequence of random bits an excess of heads.

The memory chip with the random bits for the experiment also contained a supply of random bits to be used for "play runs" which were not to be counted. The subject was encouraged to begin each session with some play runs and to postpone the test session is the play-run scores did not look encouraging. The subjects could spread out the tests over as much time as they wanted.

At first start of a test run, the subject entered a run number, specifying the section of pre-recorded events to be used. At the end of a run, the score was displayed to indicate clearly how well or how poorly the subject had succeeded. The small computer was easily portable so that the subject could work at home.

The different blocks of the experiment used four different kinds of feedback display. All these displays had been tested in earlier experiments. A "pendulum display" showed a pendulum swinging with randomly varying amplitude on a TV screen. The PK goal was either high or low swing amplitude. A "needle display" presented each head or tail by the lighting of a red or green lamp. Finally, a "click display" made each head or tall appear respectively as a pair of clicks either in the center of the head or on the sides.

To evaluate the results, let us first look at the scores of the two subjects individually. We measured the success of the subjects efforts in terms of Z-scores, closely related to the statistical significance levels.

When the first subject (the martial arts teacher) acted as first observer, his PK effort was significantly successful with Z(first) = 3.11. But when the runs had been preobserved by the fish, the subject's success rate dropped to a nonsignificant Z(last) = 0.01.

For the second subject (H.S.), the corresponding values were Z(first) = 1.8 (moderate success) and Z(last) = -0.86 (slight, nonsignificant missing).

Combing the results of the two subjects, we found significantly positive scoring (Z=3.47, odds against chance of 1000:1) for the runs without fish observation but chance scoring (Z = 0.04) for the runs that had been previously displayed to the fish. The difference was statistically significant at the 1% level.

These interesting results might have to be viewed with some caution. With the limits of PK not yet well understood, the possibilities of unwanted experimenter PK effects distorting the outcome of a particularly exciting experiment cannot be quite dismissed. Such doubts can only gradually disappear as the work is continued and replicated.

Copyright © 1998 by Guy L. Savelli. ALL RIGHTS RESERVED.
Except as otherwise provided by law, this writing may not be produced in whole or in part, in any manner.