Published in William G. Roll and John Beloff (editors), Research in Parapsychology 1980 (Metuchen, NJ: Scarecrow Press, 1981), pp. 45-47
(This article is cited in Brian Martin, "Psychic origins in the future", Parapsychology Review, volume 14, number 3, May-June 1983, pp. 1-7.)
Donald and Martin (EJP [European Journal of Parapsychology], 1976, pp. 17-36) set forth a theory for the physics of psi, based on time-symmetric thermodynamics. Their thermodynamic equations are solved for both positive and negative values of time, thus showing future events influencing past events.
The theory makes many predictions about psi. Some are strongly supported; some untested; none disconfirmed. One untested prediction is that "results of psi tests using truly random events will be more significant than those using pseudo-random events" (p. 33). We performed an experiment to examine this.
Methodological decisions were: (a) to control individual differences in psi ability by using a within-subject design; (b) to determine order of conditions randomly; (c) to control conscious preference effects by keeping both experimenter and subject blind to the target condition (though obviously extrasensory knowledge could not be shielded); (d) because our random events generator (REG) lends itself readily only to precognition or PK and because a fully pseudo-random series permits clairvoyance, to initiate our pseudo-random sequences from random choices; (e) to analyze hits by subject preference and run order as well as by target type.
An REG designed and constructed by Edwin May (and generously donated by Charles Honorton) produced random binary sequences. The subject throws a switch; the instrument displays the outcome.
A table of random numbers provided the odd and even digits for pseudo-random sequences. Each subject drew slips of paper from concealing envelopes to designate page, column, and row for entering the table.
Subjects were 50 volunteers, chiefly college students: 48 sheep and two supersheep. Thomas Krusz tested seven and G. R. S. tested 12 of the first 19 subjects; R. B. tested the last 31. Experimenters' scores showed no significant differences and were pooled.
The experimenter introduced the procedure by saying that it was like guessing how a coin would fall, and tossed a coin to illustrate. Without observing its fall, the experimenter pushed it beneath some papers. It remained concealed until all calls were completed. Its toss determined whether the first two runs were for random or pseudo-random sequences. The subject checked a supersheep-sheep-goat questionnaire. Sheep or supersheep then made 50 binary calls, rested quietly until they declared themselves refreshed, and made 50 more calls.
The experimenter next examined the coin. Targets were determined first for the first two runs. For pseudo-random targets, the experimenter explained how the table showed odds or evens, the subject selected slips of paper to enter the table, and 50 successive digits specified the targets. For the random sequence, the experimenter took the subject to another room and demonstrated the REG. The subject pressed its switch to determine 50 targets, which the experimenter recorded on a separate page. Forty-six subjects were asked to express preference as to method. All subjects saw their scores before the session ended.
There was insignificant psi-missing for both random (deviation = -45) and pseudo-random (deviation = -40) sequences. Total scores were significantly low (t = 2.72, 49 df, p <. 01, two-tailed).
Only two subsets of scores were significant, both for the subjects where pseudo-random sequences came first. The eleven subjects who preferred random sequences psi-missed for random sequences on Run 3 (t = 3.50, 10 df, p < .01, two-tailed). The ten subjects who preferred pseudo-random sequences psi-hit for pseudo-random sequences on Run 1 (t = 2.57, 9 df, p < .05, two-tailed).
To find if subjects were self -consistent, a routine (but not a formally planned) analysis was performed: correlating scores within each pair of runs. Runs with random sequences showed a low but significant positive correlation (r = +. 28, 48 df, p < .05, two-tailed). The correlation for pseudo-random sequences was r = +. 02. The difference between correlations was not significant. The correlation between totals for random and pseudo-random sequences was significantly negative (r = -. 29, 48 df, p < .05, two-tailed), but since the correlation between first and second halves was -. 27, this may be only an order effect.
When we evaluate the data, both subjects' typical reactions and the psi-missing for total scores permit one trivial conclusion: the task was dull.
For comparison of random with pseudo-random sequences, preplanned analyses showed no major differences. Each type had one significant score (when not corrected for selection). Neither showed overall significance. Random sequences perhaps showed a faint, nonsignificant trend toward more extreme scores. It is striking, however, that random sequences showed a significant positive correlation between successive runs, while pseudo-random sequences did not. This gives Donald and Martin's theory some support and deserves further testing.
For subsequent research on random versus pseudo-random targets, we recommend continued attempts to control experimenter as well as subject preference, and order effects, but a change to a more interesting procedure.