| Abstract/Results: | RESULTS:
BACKGROUND:
Recent metanalyses have shown that psi-research protocols coupling subject optimization (or "noise reduction") procedures with a free-response testing approach produce a moderate degree of replicability and adequate effect-sizes. However, these protocols are time- and resource- intensive; a single trial, or data point, can take one hour or more. Thus, in a field of limited resources, they are not well suited for process-oriented research.
The Sharefield project explores an alternative approach: an automated experimental protocol that combines optimization procedures with forced-choice testing, involving higher data collection rates. The project's long term goal is to develop a readily replicable testing protocol that produces superior effect sizes and is useful for process-oriented research.
OBJECTIVES:
1. Develop and perfect a novel automated testing approach combining forced-choice testing with subject-optimization procedures, so as to obtain a flexible and powerful psi-research tool.
2. Conduct a telepathy experiment based on this approach, to assess its viability and determine improvements. Integrating a comparison between subject-optimized conditions (OS) with Non-Optimized conditions (N-OS), the experiment tests three hypotheses:
I) the trial effect size for the OS condition will be significant
II) the trial effect size for the OS condition will be superior to that obtained in the N-OS condition
III) the OS session effect size will be superior to 0.135, the ES established for the Ganzfeld.
RESULTS:
Main Analyses
For the Optimized-Sessions (OS), we obtained 253 hits out of 500 trials (Z = 0.268, E.S. = 0.012, P=N.S.). Hypothesis I was not confirmed.
The difference in trial effect sizes between OS and N-OS conditions was ES = 0.099 (Z = 1.35, P= 0.09). Hypothesis II was not confirmed.
The OS session effect size was ES = 0.0537. Hypothesis III, was not confirmed.
Secondary Analyses
The null results observed in the Main Analyses may have been due to a combination of psi-hitting and psi-missing, rather than an absence of psi. To test this, we assessed hit rate variability by grouping trials according to three factors that could produce significant variability in scoring: participants, targets and trial position within the session. Hit rate distributions for these three factors, for both OS and N-OS conditions, were tested against the null hypothesis using a Pearson chi-squared goodness-of-fit test and a Monte Carlo test of the variance. In this analysis, we obtained significantly low probability values for two of the six OS tests and none for the N-OS; a Monte Carlo simulation estimating the probability of finding 2 of 6 tests with P-values of 0.03 and 0.02, or less, is significant (P= 0.012; 90% confidence interval of 0.007, 0.017). Thus, scoring specifically during the OS trials was anomalously variable, suggesting that the Optimized condition elicited psi-hitting / psi-missing patterns rather than no psi at all.
CONCLUSIONS:
For this first study, we are encouraged by the successful integration of optimization procedures in an automated multiple-trial protocol, and by the positive feedback given by participants. While the main hypotheses were not confirmed, the post-hoc analyses provided some evidence that the Optimization protocol did elicit psi functioning, albeit in an unstable manner. Based on the quantitative and qualitative results, we are now working on improvements for the next Sharefield experiment.
|