Abroved For Relea RDP96-00788rO 13002000014 We S, '~f 5 August 1980 Revised 2 April 1981 FEASIBILITY STUDY ON THE USE OF RV DETECTION TECHNIQUES TARGETS~~ TO DETERMINE LOCATION OF oved ForPxlJeasev28W08W r--) 326-6-200 - Cable. SRI IN 6-j P#P0b~0,qQWkW%'j30020L000f-6 Approved For Release 2 inr r- DP96-00788r001 300200001-6 CONTENTS LIST OF ILLUSTRATIONS . . . . . . . . . . . . . . . . . . . . . I OBJECTIVE . . . . . . . . . . . . . . . . . . . . . . . . . II INTRODUCTION,AND BACKGROUND . . . . . . . . . . . . . . . . 2 A. Location of Unknown1111010MTargets . . . . . . . . . 2 B. Remote Viewing (RV) as a Location Technology . . . . . 2 C. Conclusion . . . . . . . . . . . . . . . . . . . . . . 5 III hIETHOD OF APPRO.ICII . . . . . . . . . . . . . . . . . . . . 6 A. Step 1--hTicrocomputer-Based Screening Traini W . . . . 6 1. Sequential Sampling Statistical Averagin,r Procedure . . . . . ... . . . . . . . . . . . . . 8 2. System Error . . . . . . . . . . . . . . . . . . 12 3. Test Data . . . . ... . . . . . . . . . . . . . . 13 4. Summary . . . . . . . . . . . . . . . . . . . . . 13 B. Step 2--Simulation Testing . . . . . . . . . . . . . . 15 1 C. Step 3--Denion-ctration-of-Feasil)ility Field Study . . . 15 IN' PROPOSED PROGRAM . . . . . . . . . . . . . . . . . . . . . 17 REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . .. . 18 APPENDIX . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Approved For Releas e 20 6-00788r001 300200001-6 9-P low aw so W low goo in 6-007 Approved For Release 200 88r001300200001-6 ILLUSTRATIONS 1 Computer Modeling Task . . . . . . . . . . . . . . . . . . 2 Decision Graph for Site Selection . . . . . . . . . . . . . 3 Average Number of Trials n 1 to Screen Positive . . . . . . 4 Decision,Graphs for Site Selections Based on the Data of Subject 1 (Table 1) Screening Study, Resulting in Twelve Consecutive Correct Selections . . . . . . . . . . . . . . 7 10 11 14 Approved For Release 2 --~D196-0078&001300200001-6 T. Approved For Release 200 07: CIA-R 6-00788r001300200001-6 or, 7 I OBJECTIVE "0 40 go The purpose of this document is to provide an outline of a program to assess the feasibility of using RV detection techniques to determine the location of argets of interest. Throughout this document the abbreviation RV refers to the tc-1,11i "rerliote viewing," not to its other use as "re-entry vehicle." Approved For Release 20 r'IA'T"-%P96-00788rOOl3OO2OOOOl-6 Approved For Release 200 ~6-0078MO130020009.1-6 II INTRODUCTION AND BACKGROUND B. Remote Viewing (RV) as a Location Technology Of particular interest along the psychoenergetic lines is a human information-accessing capability that we call "remote viewing- (RV). The RV phenomenon, under studv at SRI International for the past nine years, W pertains-to the ability of certain individuals to access and describe, by means of mental processes, information blocked from ordinary perception 00 by distance or shielding, and generally believed to be secure against sucli access. This has included the ability of subjects to view remote geographical AN locations given only -eo-raphical coordinates or a designated person on w1iom to target. In problems of the location type (which have not been addressed in any detail in former programs) the general prospect of a continuum of 2 Approved For Release 26~0_ 96,00788r001300200001-6 A. Location of Unknown rgets The RV abilities of several subjects have been developed to the point where thev can describe--often in great detail--geographical and technical material such as natural formations, roads, buildings, interior laboratory Approved For Release 200010 - 07BBrOO1300200001-6 moil 17 _770 Possible locations can often be reduced to that of a set of discrete possi- bilities. This is because, tor example, only a finite number of awl sites ;are available, or because specifying one of a number of grid squares is sufficient to define location. If a location W1 task can be so defined (to be one of a discrete set of possibilities), then a detection method can be designed around one of the standard formats for RV testing, a statistical form of shell game which is a direct analog of the discrete location problem. One of the standard formats for RV testing is a computerized form of ll" f th h hi h i t l t l di s c e game w s a rec ana og o e targe ocation situa- tion. The testing procedure addresses the basic problem.of choosing, by - t" chni ues RV t o f b f ibl lt e correc answer r m among a num er o erna- e a poss , a q tives. An example is provided by an electronically-automated screening t d i d t Ch t b SRI lt l T t S s u v carr ar e ou consu an es v ar ubjects were asked to . determine which one of ten possible positions an a circular display had ' b d t een esigna ed as an active target by the electronic test device s random number generator. .3 From an unselected population of 2000 university students participating in a mass card screening program, seventy of tile better subjects accepted an invitation to be further screened using tile automated electronic testing system. Of these, ten were finally chosen to participate in a formal study involving 500 trials each. The results obtained with these ten subjects are shown in Table 1. It is seen that five of the ten subjects scored significantly above chance, all in the range of 1.5-2.5 times chance expectation. Tile best subject averaged a UW 24.8'~ hit rate (-2.5 x chance) over the 500-trial seqtience; tile probabilit-, of such a result or better occurring, bv chance is only p = 2 X 10-28. W Furthermore, as good as these results are, the potential utility of such results can be further enhanced by the use of error-correcting statistical averagintr techniques. Such techniques have proven themselves 3 Approved For Release 20 _'_~,96-0078MO1300200001-6 Approved For Release 2 C -:,796-0078MO1300200001-6 r~L- Table 1 ELECTRON I CALL Y-AU TOMATE D SCREENING STUM Probability of Obtaining Hit Rate Sucha Result by Chance ubject (10-, Expected)(one-tailed) 24.8'-( 2 Y 10-28 2 2 0 6;-( 1 Y 10- 14 3 16.2'-f 2 x 10- 6 4 16.0~1 4 x 10- 6 5 15.6(-( 2 Y 10- 5 6 11 .8r,* nonsignificant 7 11.4r; nonsignificant 8 10.8r: nonsignif i cant 9.4~ nonsignificant 9 10 7.V nonsignif i cant capable of amplifying even small statistical advantages to arbitrarily- high-accuracy results. To cite an example, Czech researcher Dr. Milan Ryzl, a chemist with the Institute of Biology of the Czechoslovakian Academy of Science, carried out an experiment with a subject. whose base performance level was that lie was generally capable of generating better than - W; hit rate targeting on sequences of random binary digits, or bits (0, 1), where chance expectation was 50',. For the purpose of showin,,'~ the power of psi enhancement by statistical averaging techniques, Ryzl chose as a task the acquisition, without error, of a 50-digit random binary sequence. The effort took 19,350 calls, WX averaging 9 sec per call. The hit rate for individual calls was 61.9c,, 11,978 hits and 7372 misses. 7 By means of repeated passes through the t, 4 Approved For Release 2C ca, --qD ~6-0078MO1300200001-6 f A-, OprIoved For Releas =MIQ7 Ld IbP96-0078MO1300200001-6 sequence and an elaborate (though inefficient majority-vote protocol, the subject was able to identify with 1001 accuracy all 50 bits. The 15 probability that he did so by chance is only one in 10 C. Conclusion Thus, data already extant from RV detection experiments indicate that (a) one target from among a number can, with some statistical advantage, be determined by RV detection techniques, and (b) the accuracy of doing so can be amplified by statistical averaging techniques. These observations thus provide a sound basis upon which to estimate the feasibility of RV detection of randomly distributed targets, and the protocols in use are essentially directly applicable in their present form. aw An increase in efficiency by a factor of about 9-0 could be expected on the basis of a statistical averaging procedtire more Optiminii than that dw used in the experiment.1 5 Approved For Release 20,(C Z_DP9~6 00788rOO1300200001-6 P Approved For Release 2000(~'^~. 0%1 A ~~Vr% -00788r001300200001-6 III METHOD OF APPROACH With regard to determining the vulnerability of targets to No RV detection, an approach that recommends itself is a gradient-scale three- step program involving (1) microcomputer-based screening/training, (2) simulation testing, and (3) demonstration-of-feasibility field study. Each of these are discussed below. A. Step 1--Microcomputer-Bas6d Screening/Training The first step of the program would involve screening./training a population of volunteers using microcomputer-based modeling of the location problem. Basically, the individuals participating as remote viewers are asked, in repetitive trials, to determine which one of twenty possible locations (schematically represented as circles on a computer- driven graphics display) has been designated as the simulated target by the computer's random number generator. The computer display is driven bv an LSI-11 microcomputer which, on a trial-by-trial basis, generates a new random display of the circles (to circumvent bias on the so part of the remote viewer due to previous choices). The individual enters his selections by button press on a hand device positioned over an X-Y Wd grid (see Figure 1, where a one-in-ten case is shown), and the computer responds by giving immediate feedback as to the correct answer (to encouraf,'-e learning). As the trials progress, the selections are computer analN-Zed on line by a statistical averaging program, the output of which indicates whether one of the possibi li ties has been chosen statistically significantly more often than expected by ciance. (In the later application phase essentially the same procedure is followed, with the circles internally 6 Approved For Release 2 ~)P 6-00788r001300200001-6 ww- -- a a d f - go 10 -a 39 39 3 FiGURE I COMPUTER MODELING TASK, The circles representing possible target locations are shown in the lower video monitor, a decision graph is shown on the upper momior. The remote viewer's choice is entered by button press on hand device posit~onecl over x-y gfid. Approved For Release 2000/08/07 CIA-RDP9.6-00788rOO1300200001-6 keyed to actual target site possibilities. The procedure differs only in that trial-by-trial feedback would, of course, not be available). 1. Sequential Sampling Statistical Averaging Procedure do An efficient statistical method for the screening./training process is provided by a sequential-sampling technique used in production- aw line quality control.a The sequential method gives a rule of procedure for making one of three decisions (with regard to each of the possible choices) following each trial, which consists of a remote viewer entering a selection: the accumulated selections have met a pre-established hit- rate criterion (decision positive); the accumulated selection do not exceed chance expectation (decision negative); continue trials (insufficient go data to make a decision). The sequential sampling procedure differs from fixed-trial-length procedures in that the number of trials required to iW reach a decision is not fixed, but depends on the results accumulated with as comprred with other methods is that, on the average, fewer trials per each trial. The principal advantage of the sequential sampling procedure decision are required for an equivalent degree of reliability. To apply the sequential analysis procedure to screening training t, ~? MW we must a priori define the hit rate we require to conclude that useful RV detection is taking place, and -what statistical risks we are willing to M accept for making an incorrect decision. To meet these criteria, sequential analysis requires the speci- fication of' four parameters to determine from which of two distributions (chance or required-hit-rate) a data stream belongs. They are: p , t lie fraction of selections of a particular target expected in the chance condition (e.-., p = 1/20 for the case under discussion); p the fraction f, 0 of selections expected in the presence of a functioning RV capability (C.tg., i = 0.125 for a 2.5 Y chance-expectation requirement, a value that might p 8 Approved For Release 20 08/07: CIA-,ROP96-00788rOO1300200001-6 Approved For Release 20001or 7-- CIA.R -00788rOO1300200001-6 7 be chosen because of Previous performance in a successful one-in-twenty task); 0!, an a priori assigned acceptable error rate (e.g. , Ot = 0.05) for . concluding that accumulated selections of a particular choice derive from the P1 (RV) distribution when in fact they derive from the p (chance) distribution (Type I error); B, an a priori assigned acceptable error rate (e.g., 0.05) for concluding that accumulated selections of a particular choice derive from the p (chance) distribution when in fact they derive 0 from the p1 (RV) distribution (Type II error). With the parameters thus specified, the sequential sampling rocedure rovides for construction of a decision ra h of th t sho wn e g ype p p p in Figure 2. The deci.sion graph illustrates the rules of procedure for -; followin makin one of the three ossible d cision ch t l ti i . e : con r nue g ea a g p test before making a decision (unshaded middle re,rion in Figure 2); d cisi iti d Fi h d i i 2) d ( e ve on n on pos a e gure upper s reg ecision negative ; (lower shaded area in Figure 2). The equations for the upper and lower decision lines are given in the Appendix. With the appropriate equations pro,--rammed into the microcomputer, the computer automatically records all data (trial number, target response pair), and displays on the video graphics system protrl-CSS on a target decision graph. A cumulative record of remote viewer selections is compiled by the computer until either the upper or lower decision line is reached, at which point a decision is made. Also given in the Appendix are the equations for the average number of trials to make decisions, positive or negative. A plot of the average number of trials to reach a positive decision for typical cases of interest is shown in Figure 3, where 5', (a'9 1) error rates have been assumed. As an example, we see that for a 2.5 Y expectation rate (k = 2.5) hitter, n 62 trials are required on the average to reach a positive decision on a one-in-twenty target. Approved For Release 20 96-00788001300200001-6 Approved For Release 2000/08/07 : CIA-RDP96-00788rOO1300200001-6 C 2. System Error The overall system error is dependent on the type of mode employed in site penetration attempts. (a) If the RV detection task is approached with a tentative choice having already been made (presumably by more conventional means), then the task of the remote viewer is to verify or reject the tentative decision as a back-up test. In this mode, only a single decision graph is plotted in the target choice of interest. The probability of error due to chance ~Pe PC ) in this case - CY, being given by the product of the probability of making a selection even though operating at chance, and the percentage of such selections that correspond to an incorrect decision: P c = (N - 1 a e, (b) If the RV detection task is approached as a blind one-in-N task one-in-20 task), the N decision graphs are plotted in parallel, one for each of the N target choices, as each selection is being made. In this case, to a good approximation the graphs can be treated in the chance condition as independent, and the probability of error due to chance (P Mi. Specifically, it is given by the product of the probability e,C of making at least one selection in the N .-raphs by chance (which is one minus the probability of making no selections), and the percentage of such selections that correspond to an incorrect decision: P , C = (N - 1 CY)N] Q N For example, with N = 20, a I"(* individual-target error rate t, (a = 0.01) leads to P 0.17, or a confidence factor 1 - P 0.83; e,c C'C this provides - a 17-fold increase in odds over the one-in-twentv confi- ,dence factor expected by chance. ~2 MW Approved For Release 20 96-00788rOO1300200001-6 Approved For Release 2000/08/07 CIA-RDP96-00788rOO1300200001-6 3. Test Data As a test of the above procedure applied to real data, the data generated by Subject #1, Table 1, were processed by passing it through the sequential analysis statistical averaging program (500 trials, 24.81, hit rate on a one-in-ten task). With the parameters set to correspond to a twi ce-chance-expectat ion requirement and 5c,,. (Cr, ~) error rates, the results are as shown graphically in Figure 4: twelve correct selections, in a row, of one-in-ten targets were made in 452 trials. Although the data was gathered under the condition that the correct answers were stored in the computer during the runs, and therefore trial-by-trial feedback could be given as the random number generator stepped through its program, the conditions are nonetheless sufficiently similar to the projected task that the results can be taken as evidence that the proposed approach is sound. 4. Summary for each run would be designated internally by the computer's random number trained by carryin-- out the task described in this section, first with trial-by-trial feedback to encourage learning, and then without feedback to model properly an application studv. In this initial phase the target generator. In the screening.'training program, participants would be screened Carried out on a large-enoug~h scale, the screening training program described in this section would provide realistic estimates of the percentage of population trainable in this task, and the levels of proficiency to which performance in this task could be developed. In a program designed to assess to its fullest the feasibility of locating .targets by RV detection techniques, it is recommended that suffi- iently large-scale screening to meet these requirements be considered. 13 'pproved For Release 21 )P96-0078MO1300200001-6 P Approved For Release 20001C_ 6-00788rOO1300200001-6 B. Step 2--Simulation Testing The Participants who emerge from Step 1 with successful performance profiles would then be asked to participate 2. For this step, in Step a model of an actual ituation with a randomone-in-twenty designated target would be constructed. The subject'saccess to the mockup during experimental runs would be by way of videomonitor,although secondary means such as maps or photographs might utilizedin later stages of be the study if appropriate. To carrv out the test, a participant (or participants) would be briefed as to the task and then be asked to proceed as in Step 1. The sequential sampling parameters in the microcomputer analysis program would be set in accordance with the performance profile established by the par-. ticipant(s) in the Step I screenin.--'training study. In Step 2 the mechanics of microcomputer recording and analysis of subject selections would be the same as in Step 1. Step 2 differs from Step 1, however, in that a participant's selection from the random circle display, internally keyed to numbered sites, cannot be internally compared to a recorded correct answer. The results generated by the participant(s) in the site selection procedure would then be tabulated and discussed,- Should the results appear encouraging, then Step 3 would'be engaged. C. Step 3--Demonstration-o.---Feasil)ility Field Study The final step in the three-step vulnerability assessment program would consist of a field-demonstration test involving,,' Data would be taken using the successful remote viewers of Step 2, bot h to determi ne am the degree of correlation between performance on the tasks of Steps 2 and 3, and also to evaluate actual performance in the field studv. 5 1 Oir Approved For Release 20 Approved For Release 200 r796-00788rOOl 300200001-6 The possibility of success in such a field study is buttressed by the fact-that the procedures described here have been used by us success- fully in an exploratory program to determine the locations of hidden material. Following a series of such tests, performance profiles for the individual remote viewers would.be computed and the overall data set would be evaluated to provide an.estimate as to the usefulness of RV techniques MW No 16 Approved For Release 2 P96-00788rOO1300200001-6 IN Approved For Release 2000/04 _76-00788r001300200001-6 11' PROPOSED PROGRAM To accomplish the proposed program, SRI proposes to provide the necessary personnel, facilities, and materials to perform the outlined work, summarized below, and to report on the results thereof. 0 Provide to the sponsor the details of the statistical package and hardware setup tailored to sponsor-designated task requirements. 0 Screen,,train a population of volunteers on an LSI-11 microcomputer-modelled location problem, first with real-time feedb:,.ck, then without (Step 1). 0 Carry out simulation tests on mock-ups of an actual tar--et situation using participants with successful performance profiles from Step I (Step 2). Carry out a demonstration-of-feasibility field study on a sponsor-designated test site of interest (Step 3). Evaluate data sets to provide estimates of: (a) Percentage of population trainable. (b) Level of proficiency to which task performance can be developed. (c) Usefulness of locating. targets by RV detection techniques. It is proposed that the above program be pursued on a three-man-year level-of-effort basis. If programmed as a two-year effort, an expenditure. of somewhat less than $200K for the first year is envisioned. An itemized cost breakdown can be provided on request. low I 17 _*i, Approved For Release 20 0/08/07::~-, 00788r001300200001-6 Woo _~j 0