The MegaREG Experiment: Replication and Interpretation

Anomalous effects of human intention on the output of electronic random event generators (REGs) have been well established at the PEAR laboratory and elsewhere. A simple model of this effect as a change in the binary probability of the REG digits would predict that larger statistical yield can be achieved simply by processing more bits. This hypothesis was explored previously using protocols ranging from 20 to 2000 bits per trial, with results that were consistent with the bitwise model, but had too little resolution to rule out many competing models. More recently, a ''MegaREG'' experiment was deployed to test this hypothesis using 2-million-bit trials interspersed with 200-bit trials in a double-blind protocol.

In the initial phase of MegaREG, the 200-bit trials produced outcomes comparable with our standard experiments, while the 2-million-bit trials produced an effect somewhat larger in absolute scale, but inverted with regard to intention. A subsequent replication phase reproduced these findings, except for statistically nonsignificant quantitative changes. These appear to be secondary consequences of a statistically significant difference between operators having, and lacking, prior experience in REG experiments, the relative proportions of which account for the differences between these experimental phases. Other operator population distinctions, such as gender, and various secondary protocol parameters, had no significant effects.

A related experiment called ''MegaMega,'' differing from MegaREG only in that all data used 2 million bits per trial, with no interspersal of a second data type, produced a reversed intentional effect of the same scale. It also displayed a significant asymmetry between the intentional runs and the non-intentional baselines, which was not seen in MegaREG.

The combined result of all high-speed experiments was an effect size per trial of
2.77 6 0.69 times that seen in earlier REG experiments, but given the larger number of bits per trial, the bitwise probability change was some 30 times smaller. The composite score for the intentional effect in high-density data across all experiments was T ¼
4.03 (d.f. . 105), p ¼ 5.65 3 10
5 (2-tailed). The causes of the change of scale, and of the inversion of sign in the effect, remain unknown. Explanations that can be ruled out with a high degree of confidence include statistical artifact, the change in the source, the use of different operator pools, and the double-blind interspersal of data types.

Testable explanations that remain potentially viable include increased task complexity, inherent timing or rate limits on anomalous functioning, and changes in the psychological environment.