Approved For Release 2 96-00787ROO0200150005-1 02911U914142L ~ VVIII IL ORD Z240-75 12 June 1975 MEMORANDUM FOR: OTS/CB SG1 I ATTENTION: SUBJECT: Evidence for Non-Randomness of "Pour-State Electronic Random Stimulus Generator" REPERENCE: OTS/CB Memorandum 075-60 As requested in the last PaTagraph of the referenced memoranduTq, we have investigated the data provided to establish evidence for randomness. The basis for suggesting non-Tindorness is as follows: Table I of the Reference provides data concerning fre- quencies of: A. Initial States R. State Transitions Since the experiment consisted of requiring the subjects to indicate the next-to-be presented state, it would seem 'most important to establish that all possible transitions occur with equal probability. To test for possible non-equality of transitions, we extracted the observed frequencies of non- identity transitions to form the following table: Yellow Green Blue Red Y - 764 765 790 G 777 - 773 863 B 776 796 773 R 787 SS2 803 - Approved For Releh"J2@0WM1W' A. t AM DWR A ~ AY0 AM& 0 0 5 - 1 Approved For Release 209912JI12 - E.IA-RDP96-00787ROO0200150005-1 WI-14 I OIL ORD 2240-7S This table can be restructured as a two-by-six table as follows: yLG VB X/R 2/R E,./ B/R R Porward, 764 765 790 773 863 773 Backward 777 776 797 '796 8S2 903 The table thus restructured brings together all possible ton-identity transitions viewed as state-paIrs. for instance, Col. 1 shows that there were 764 trans t ons Yom the yellow state to the green state and there were 777 transitions from the green state to the yellow state.. Under the hypothesis that all state transitions are equally probable and equally affected by chance the Observed frequency with which forward transitions occur should be unrelated to that with which backward transitions occur in the Same Pair. This condition is not met. There is a very Strong relationship between the observed forward and back- ward transition frequencies. The coefficient of correlation between frequencies for these two directions, computed across &11 six possible non-identity transitions is .93, (p< .01) (see attacbed graph). This finding shows that there were, in fact, systematic pair-wise blasts associated with the electronic processes by which the transitions were selected. The finding that the forward and backward transitions are closely associated with respect to joint probability of occurrence susgests that they can be considered as having been drawn from the same population. To test this, we computed the forward and backward mean and the Standard Deviation (SD) of the observed frequencies. They are# Mean SD forward 788 37.9 Backward 798 28.25 The standard error of the difference between these two means is IS.S9 while the difference between ther, is only ten; clearly Approved For Release 200C~19,0/12_ a =-PI,9,6-00787ROO0200150005-1 UVIII IUL-11 I in-L- Approved For Release 2Q"C&4ffi=W-RDP96-00787ROO0200150005-1 UWAV I TUETMAL ORD 2240-75 those data may be merged. Merging them, provides the following table: Observed Frequency Expected Frequenc Under Null Hypothesis Chi Square YIG )2R G/B R/R R Total MR !L 1541 1541 1S77 1S69 171S 1576 9519 1586.S 1SMS 1S86.5 IS86.5 1586.5 1SMS 9519 1.305 1.30S .0$7 .193 10.408 Total Chi Square - 13.337 df-5 p-.02 .069 In the above table it can be seen that the large excess of observed transitions involving the red-gTeen pair Is significant at the .02 level. Inspection of the observed frequencies reveals that there are almost ten percent more transitions involving the red-green pair than the average of the other five possible no-n- idontity transitions. Those results su gest that adopting (fOT whatever conscious- or unconscious reasons a strategy of "When green,, press red, when red, press gTeen and, OtherWige USO the 'pass' button as much as possible" will increase one's hit.score. Usin an instrument with the above-described characteristics ana strategies such as this is certain to-produce "statistically significant" results, given enough trials and the asstuaption of random transition robabilities. Other biases also exist which could form the Uses of other enhancing strategies but the above discussion would seem adequate to establish the existence of non-randonness which we have suggested. The report available to us contains data only upon one test of one instrument. It must, therefoye, be Assume4 that the other instruments demonstrated non-randor characteristics of a similar nature. Further, the report does not reveal which subject used 3 Approved For Release 2 6-00787R000200150005-1 UWJ_A1 Approved For Release 2oodMIK MAJOglo-00787RO00200150005-1 M EV so- ORD 2240-75 which instrument so we are unable to ascertain whether or not subject number 2's results could be due to the effects discussed above, but the magnitude of the effect is adequate to explain the results If one assumes the adoptlonof a selection strategy which "capitalizes" upon the non-random characteristics which are demonstrably present. SG1 I LSR ORDMD 1k Attachment Distribution- Original & 1 - Addressee, w/att -1 - LSR File, w/att I - LSR Chrono, w/att I - ORD Chrono, w/att LSR/ORD/DD/S&T~~:mav-3191 SG1 1 4 Approved For Release 2000/08/10 : CIA-RDP96-00787ROO0200150005-1 41" 1% 1 r-1 QrAITI A I ki UTTI i L.., L-T-rrrrm