Operator-Related Anomalies in a Random Mechanical Cascade
B. J. Dunne, R. D. Nelson, R. G. Jahn

Princeton Engineering Anomalies Research, Princeton University,
Princeton, NJ 08544

Experiments with a "Random Mechanical Cascade" (RMC) apparatus have
yielded anomalous results correlated with pre-stated intentions of human
operators. Based upon a common statistical demonstration device, this
machine allows 9000 polystyrene balls to drop through a matrix of 330
pegs, scattering them into 19 collecting bins with a
populationdistribution that is approximately Gaussian. As the balls
enter the bins, exact counts are accumulated photoelectrically,
displayed as feedback for the operator, and recorded on-line. Operators
attempt to shift the mean of the developing distributions to the right
or left, relative to a concurrently generated baseline distribution. Of
the 25 operators who have completed one or more experimental series with
this device, four have achieved anomalous separations of their right and
left efforts, and two others have displayed significant separations of
either their right or left efforts from their baselines. The overall
mean difference of right versus left efforts concatenated across the
total data base of 87 series (3393 runs), has a probability against
chance of <10-4, with 15% of the individual series significant at p
<.05, and 63% conforming to the intended directions.

The concatenated results display a stark and curious asymmetry, in that
virtually all of the right vs. left separation is provided by the left
vs. baseline separation. This pattern also appears in the data of
several individual operators, and is not attributable to any known
physical asymmetry in the experimental system. In addition to the
systematic asymmetric deviation of the distribution means, cumulative
excesses in the variances of the left and right distributions relative
to baseline are also observed, progressing to statistical probabilities
of .003 in the left efforts, but only .2 in the right. More detailed
study of the individual bin population patterns reveals that while most
of the bins contribute to the overall mean shifts and variance changes,
those on the outer portions are more influential than those near the center.

Operator achievements tend to compound marginally but systematically in
cumulative deviation patterns characteristic of the particular
individuals and, in several cases, similar to those produced by the same
operators in microelectronic Random Event Generator (REG) experiments.
Within these characteristic patterns of achievement, some operators also
show sensitivities to secondary experimental parameters, such as
instructed vs. volitional establishment of the intended directions, or
the presence or absence of feedback displays. Other successful operators
seem insensitive to such options.

Two major protocol variations have been explored, one employing remote
operators, the other, multiple operators. In the former, operators with
well-established performance in local experiments attempt to influence
the bin distributions from remote locations up to several thousand miles
from the laboratory. Significant results are again obtained that are
quite similar to those of the local experiments, with the exception that
the overall right and left distribution variances are smaller than those
of the baseline. In the multiple operator experiments, early results
show little resemblance to those achieved by the participating
individuals alone.