RPKP Update No. 8

January 31st, 1997
by John Walker
This was originally sent out to members of the RPKP mailing list. If you would like to receive future updates, please see the instructions explaining how to subscribe to the mailing list.

RPKP Experiments: Pilot Phase Begins

Since its inception, the goal of The RetroPsychoKinesis Project has been to make available, on the World-Wide Web, experiments which explore whether subjects can influence, with statistical significance, prerecorded but unobserved streams of binary data from a hardware random number generator. The Web is an ideal medium for experiments of this kind since volunteer subjects with Web access can participate from anywhere in the world, setting their own schedule, without the need to transfer data on physical media such as floppy discs. The experiment programs are written in Java, which allows them to run on any computer with a Web browser that supports Java applets.

Pilot Phase

The first experiments are now available on the Web in a "Pilot Phase" release, announced only to RPKP mailing list subscribers; they are not yet linked to the RPKP home page nor anywhere else at this site. Experimental runs by subjects during this phase may have to be excluded from subsequent statistical analysis should the pilot phase reveal a defect in the design or interpretation of one or more of the experiments. (In no case will data be destroyed--all data are tagged with the version of the experiment which created them, allowing analysis programs to exclude data from experiments later determined to have been flawed.) Given the very simple experimental protocol, I believe the probability of error is small, but genuine confidence comes only after real-world testing and examination of the implementation by others. The pilot phase is intended to build confidence in the experiments before publicising them more widely.

Outline of the Experiments

Retropsychokinesis experiments are performed in the following phases.
Random Data Generation
A stream of random bits is generated by the HotBits radioactive random number generator and stored in an inventory on the www.fourmilab.ch server. Extensive precautions are taken to ensure data in the inventory are not examined before being delivered to a client machine, and to prevent a sequence of data being delivered to more than one client. An inventory of up to two million random bits is maintained, one million on the main server, and a second million on the machine to which the Geiger tube is connected.

Experiment Request
A volunteer subject, visiting the RPKP Experiments page, decides to perform an experimental run. The subject chooses a visual feedback program, a goal (whether attempting to maximise or minimise the number of one bits in the sequence, expressed in terms of the display produced by the visual feedback program), and whether the run is a Demo (pseudorandom bits generated locally on the subject's computer are used to demonstrate the visual feedback program), Practice (data from the HotBits generator are used, but the run, though recorded, is marked to be excluded from analysis of the subject's results), or Record (the run is logged under the subject's identity [E-mail address or an arbitrary "handle" for subjects who wish to remain anonymous]).

Experiment Delivery
For Practice and Record runs, 1024 bits are withdrawn from the HotBits server, and an entry is made in the RPKP Experiment Log recording the experiment request parameters and the random bits withdrawn. (Demo runs use pseudorandom data generated on the subject's machine and do not consume HotBits nor make entries in the log.) An experiment Web page is dynamically generated and returned to the subject, incorporating the chosen visual feedback program (a Java applet) and, for Practice and Record runs, the random bitstream, passed to the applet as a parameter.

The Subject Runs the Experiment
In the absence of retropsychokinesis, the outcome of the experiment was predetermined long before, when the random bits were generated and placed into inventory, and will obey the probabilistic expectations for a random bit stream. Claims of retropsychokinesis assert, however, that certain subjects, by attempting to influence a feedback display, even if driven by prerecorded data, can create statistically significant excursions from chance. Having received the random bits and the Java applet to express them graphically, the subject attempts to bias the data toward an excess of ones or zeroes as indicated by the visual feedback program.

Immediate Result Feedback
At the end of the run, after all 1024 random bits have been expressed by the visual feedback program, the subject is shown the score (excess of one or zero bits, according to the chosen goal), and the statistical significance of that excess in a single run of 1024 bits, expressed as how often frequently a run with the score obtained will occur due to chance.

Historical Result Feedback
A subject can review the results of all runs made to date in Record mode and in Practice mode if a user identity was entered.

Database Retrieval
Anybody who wishes to analyse the results of the RPKP experiments can retrieve the database of all experiments to date, either as an interpreted HTML document or as a raw ASCII file including the random bit streams in hexadecimal form. Preserving the entire bit stream in the log permits others to perform their own statistical tests on the raw data. To preserve the privacy of subjects, their E-mail addresses or "handles" are replaced with automatically assigned subject numbers in these reports. Subjects can identify their own experiments by entering their identity in the database request, which will cause their own runs to receive a subject identification of "Me".

Visual Feedback Programs

All RPKP experiments attempt to bias a sequence of 1024 prerecorded random bits toward an excess of ones or zeroes. Three different visual feedback programs present these data to the subject; others can be implemented based on a template downloadable from this site. Please visit the experiment selection page to see the feedback programs currently available, and the template experiment if you are interested in developing your own visual (and audio) feedback programs.

Experiment-Related Resources

All documents related to the RPKP experiments are linked to the Experiments Table of Contents. Included are an Introduction to Probability and Statistics applicable to experiments such as these (I'm not at all happy with the state of this document, and expect to refine it over the next couple of weeks; reports of errors and suggestions for improvement are welcome. After working on this document for most of the last two weeks, I decided that since it is presented solely for background and plays no part in the operation of the experiments, there was no reason to delay the pilot phase of testing the experiments pending its perfection.)

Also available is a JavaScript Chi-Square Calculator that numerically evaluates the chi-square probability distribution for a given chi-square sum and degrees of freedom.

Source Code Availability

In any search for apparently anomalous phenomena, it is essential that experimental error and fraud be excluded as causes of any significant results which may occur. This can be done only by allowing others to examine the experiment design and operation, and to replicate the experiment themselves. All of the software developed for the RPKP experiments has been placed in the public domain and can be downloaded in source code form by anybody. Experimenters can establish their own RPKP Web sites, ideally using their own random bit generators so as to be completely independent of the experiments run here.

RPKP Site Speedup Soon

In mid-January, the www.fourmilab.ch site which hosts RPKP experienced a sudden doubling in the number of accesses to about 38,000 per day, which rate has persisted. This completely overloaded the 128 Kb Internet connection, and has resulted in slow response even at non-peak periods. The Internet connection will be upgraded to 256 Kb the week of February 3rd (unless something unforeseen happens in the process of installing the new line), and I expect the increased bandwidth to greatly reduce the delays and attendant frustration.

In the longer term, this Spring or early Summer a mirror of www.fourmilab.ch will be established in North America, with a 1 megabit per second (T1) connection close to the Internet backbone. When that site comes on line, requests originating in the Western Hemisphere will automatically be forwarded to the mirror site and follow-up requests served directly from there. Given the geographical distribution of Web accesses, I expect the mirror to put a permanent end to slow response.

When running the experiment programs, you may encounter a long delay the first time you try each experiment. A Java program is typically broken up into a number of ".class" files, all of which must be downloaded individually before the program can run. This takes much longer than downloading a single file the sum of the lengths of all the separate parts. The Java developers recognise this problem and, in Java 1.1 provide a way to bundle all the pieces of a program into a single compressed ".jar" file. Unfortunately, the vast majority of Web browsers in users' hands today do not accept Java applets in this form, so there is no alternative to patience: in the short term as the applets download, and in the longer term until ".jar"-aware browsers become the norm.


Experiments Table of Contents

RetroPsychoKinesis Project Home