68 Papers collapsed across time of night within stage (e.g. , early and late REM awakenings averaged to give one REM value), no significant differences were found among waking, REM, and Stage 2 trials for any PK task. Since the afternoon and nocturnal trials did not vary with the presumed underlying circadian temperature curve, it would appear that PK performance was not affected by the diurnal sleep/ wake rhythm. Control Trials (PKC). Data from the control trials were also analyzed to check that the RNG did not deviate from randomness even after several hours of ntinuous operation. The overall mean Cl) of these trials falls exactly alchance level, and an ANOVA finds no effect of gender, stage o time of night (the latter eep' or d would be expected if RNG bia develpe ov* the course of the long experimental session). 11 wever, g-nificant y (p = .052) stage-by-time in r_~tion s a m' Pna'l 00 te ufac d T "ts precisely a) 0 the interaction predicted in H thesis 3, ut to find it in the 0 control data rather than in the perime al trials is disconcerting. Explanations relating to observer effect r RNG bias are rejected (for reasons that space does not ermit etailing here). Since these trials were the last element in a v ry ng and demanding procedure, 04 a) the possibility remains that t~Wds in r ction may be a "release-of- m ter, effort" effect on the exper=s, art. a) a- Discus *on C1 The results of this study d not pport the predictions made y at the outset. No systematic. dif , difi rences ere found between REM and Stage 2 PK trials nor betwe n the wa _ng and sleep conditions. The time-of-night variable emer ed only in he correlation with SSS m'i r, Ln ratings for the PKR trials and interactio with the gender vari- T_ able for the PKE trials. No d' ect evidence as found for a "release- 00 of-effort" effect after overt t als with immedi e feedback. The major finding in t s study was an in raction between C14 stage of sleep and gender fo the experimental t 'als only. This 4) effect had not been 'e, and is somewhat su rising, since Predict ,ars no relatp M gender often be nship to psi perform e. Due to the W small sample si 'he 'enp oup (N 10) tl-ds nding should be replicated bze in t~ efore definitive. 0 A major methodologi al problem in this experimen is that the LL PK trials were scheduled considerable time after awak ing from 'D 4) the preceding sleep -tage/~due to a conflict with other e erimental u > procedures. By the trials took place, the s ject had 0 1 t'm the P &_ been awake and engagedlin v rious cognitive tasks for ap oximately CL CL half an hour. For this easo: it would be more correct to call the nocturnal trials "waking/PK" rather than considering them as strong- ly influenced by the prtceding stage of sleep. The failure to find a stage difference betwten REM and Stage 2 supports this view. Psychological Correlates 69 What is surprising is the failure to find differences between the afternoon control and the nocturnal PK trials, since some diurnal variation in performance was expected. (However, it may be noted ected a 'p that in this same ex riment, performa ce on such cognitive tasks erform a c rimen r' n t,__ as digit span, anagra , reaction tim , and several perceptual i time tasks also did not diffe between d time and nocturnal sessions; ffe be n /d in this context the PK d a are a least consistent with the pattern PK da least of performance on other t ks.) thert \ In conclusion, t e2%iment attempted to study PK per- ~t f I _e_,um ably) powerful physiological vari- formance as a functior ables of sleep stage and ime of ight. C? However, overall differ- / a e nd gh g ences between REM an non-REM wakenings, or between early and REM an M late arousals, were t found for a y measure.0 For methodological _ere tfou r-r a y 0 reasons this study annot be regar d as 00 a definitive test of these study -notI gard a) th as variables. Repli tion with PK as th primary focus of the experi eph tinw th 0 .I. in _10, K ment, with PK ials in closer proximity to the preceding sleep a stage, would required before any firm 1~_ statement can be made. 0 u eq redt 0 C14 a) BUILDING A PK TRAP: THE ADAPTIVE TRIAL SPEED METHOD J.C. Jacobs, J.A.G. Michels,t B. Millar,t and M-L.F.L. Millar-De Bruyne (Synchronicity Research Unit, P.O. Box 7625, 5601JP Eindhoven, The Netherlands) Psi can be described as altering the output probabilities of a random number generator (RNG). It does Ln not matter whether the RNG is electronic, mechanical, or a living system. This view is 00 developed particularly in the observational theories. Most lucid among these is Schmidt's mathematical model of psi. The basic idea is that the output probability of an RNG Q is changed by its connec- tion to a so-called psi source. Feedback 04 is crucial. 4) -dn the area of altering probabilities Another view of psi witl M has been expressed by Breederveld (Research Letter, Parapsychol- ogy Laboratory, University of Utrecht, 1976 [No. 7), 1-9). Psi has a tendency to cluster into successful and unsuccessful periods: psi can be seen as cyclical. Breederveld 0 suggests, further, that this clustering is a reflection, in the LL outside world, of the internal psychological states of the observer. > 0 Schmidt has discussed the possibility that the psi source fluctuates in strength. But the original CL observational theories of 0- Walker and Schmidt have little to say about what observation is. Recently, however, Millar (RIP 1986, 107-108) has proposed the Quasi-Motor Model, which explicitly deals with the observer's psy- chology. Feedback to the psi source is preprocessed by part of 70 Papers the normal human information - processing system. The preprocess- ing "program" may bring the external world into congruence with the observer's internal world. Within this general theoretical framework, a major project on cyclical psi has developed within our research unit. Breederveld first devised a method to trap periods of psi-hitting, based on a stopping strategy. In his hands, this initially gave very promis- ing scores. Later experiments lend some support to the view that it is a method for obtaining extrachance scores. However, the re- sults are not reliable and the efficiency is very low. If the period Cl) of the psi cycle were irregular and only sometimes greater than the duration of the experimental set, such deficiencies would be expected. Jacobs next devised a novel procedure, the adaptive trial 00 speed (ATS) method, to trap even relatively short periods of psi- a) 0 hitting. The basic idea is that trial speed is increased after a hit 0 I- and decreased after a miss. The intention is that the sample fre- 0 quency locks in to the psi cycle: during periods of psi-hitting, 0 0 more trials are performed than in missing periods. This should of lead to extrachance scoring in PK experiments. C14 a) This paper deals with explorations of the effectiveness of Q Jacobs' ATS method. Our unorthodox policy is to use self-test ex- periments. In our view, the classical multisubject psi test is often dominated by experimenter psi: it is, thus, merely a disguised ver- sion of the self-test experiment. For security reasons, such a de- sign is frowned upon by the Rhinean school. But if results look promising, more security -oriented work can follow. We report here U the first 8 of a planned series of 12 experiments; 4 different subject/ experimenters carried out 2 experiments each. 00 0 Procedure A computer/RNG set-up was used. The RNGs are units from C14 our standard noise-based series. Philips P2000 computers were em- ployed. The program is in BASIC, with linked machine code rou- M tiries. The standard experiment consists of 100 runs, each of 500 4) trials. A trial involves sampling the RNG, w1rich produces 0 or 1 with equal probability; the number generated is displayed on a TV monitor. The I output is the target. After a hit (1), the time to 0 LL the next trial is shortened by one step; a miss results in an inter trial interval that is one step longer. There are 12 speed steps in 4) > -le range 3 s to 100 ms. As a hedge against possible bias, the 0 output of the RNG is inverted in a pseudorandomly chosen 50 of L_ CL 100 runs. The results for each run are automatically recorded on CL digital tape for later analysis. Each subject /experimenter worked in his own home, at his convenience; on average, 5 to 10 runs, taking half an hour to an hour, were performed per day. Psychological Correlates 71 Results, Refinement, and Further Results In the first experiment Jacobs obtained 303 hits above mean chance expectation (MCE) in 50,000 trials: Z = 2.71, p < .003. Michels achieved an almost identical significant score: 304 hits above MCE, Z = 2.72, p < .003. In contrast, the Millars obtained only chance scoring (for Millar Z = -1.53 and for MiIJar-De Bruyne Z = ~0.66). In round 2, the Millars' riull results were replicated (for Millar Z = 1. 30 and for Millar-De Bruyne Z = 1. 44). For this work a fairly extensive assortment of separate con- trol experiments were performed. These gave the system a clean bill of health; they were, however, time consuming. Consequently, at this point, a methodological refinement was introduced ("double tick" sampling). For every trial two random bits were taken in quick succession. One of these was used as the experimental (E) trial proper, while the other served as control (C). Which of the bits became E was determined, on the fly, by a pseudorandom ser- ies. E and C were thus identical except that only E trials were presented as feedback to the subject. With t1ris improvement, Jacobs and Michels each ran a second experiment. In this case Jacobs' results (and controls) were at chance level (Z = 0.93). But Michels again produced a significant outcome (319 hits above MCE, Z 2.85, p < .002): the concurrent control trials were at chance (Z -0.30). Discussion and Conclusions The subjects scored true to type. The enthusiastic o -5- tor of the ATS method, Jacobs, succeeded in his first experiment but obtained only chance scores in the second. The Millars ten- dered support for the binomial theorem. Michels (JSPR, 1987, 119- 129) has recently reported a highly significant series of self-test PK experiments, based on a stopping strategy. Here he continued his unbroken run of virtuoso performances. In the midst of such subject variability, is it possible to say anything about the effectiveness of the ATS method? Chance re- sults were obtained by both Jacobs (Z = 0.64) and Michels (Z = 0.24) in separate experiments with fixed trial speed. However, it would be naive to put much weight on the results of this "same subjects" design; both sophisticated subject /experimenters had an investment in the superiority of the ATS method. Less direct but more satisfactory is the difference with previous work. Using Schmidt's psi quotient, Michels' previous highly successful experi- by comparison: the current work is about five times as ments pale efficient in the sense that only one-fifth of the number of days is needed to achieve the same significance. This affords some sug- gestion that the ATS method really works, at least for Michels. C,) 0 00 01) 0 0 1~_ 0 W C*4 a) P_ 0 9 (D a) L) Ln 00 0 0 0 C14 4) U) M a) 2 &_ 0 LL '0 4) > 0 L_ CL CL < 72 Papers Psychological Correlates 73 The ATS investigations have, as spin-offs, shed light on Michels' abilities. The work reported here shows that he is not limited to one restricted experimental procedure. In the current experiments, too, the timing of RNG sampling is determined auto- matically: there is no sample switch, as in the earlier experiments Consequently, skilled timing taking advantage of hypothetical peri- ods of local nonrandomness is ruled out as a viable explanation of Michels' success. In Experiment 2, with built-in concurrent controls, Michels' scoring was restricted to E trials, while the unseen C trials were unaffected. This strongly suggests that the effect is due to PK ~L la. observational theories. In the separate control experiment run by Michels to match his first ATS experiment, it was noticed that the number of hits (50,000) is exactly MCE. This seemed too good to be true (p < .008 [post hocD. The control (simulation) runs in Michels' pre- vious work (JSPR, 1987, 119-129) were therefore examined: in each of the three simulations, the results were also suspiciously close to MCE. For Michels, then, it seems that a separate llcon- troll' experiment may be just another kind of PK test, in which the ,goal is to score as close as possible to MCE. Jahn et al. (RIP 1985, 9-13) have reported similar findings, Continuing this line of thought, the striking correspondence between the first experiments of Jacobs and Michels was examined. Not only the total scores but also the time course were significant- ly alike. These observations suggest that Michels may have used his psi abilities to copy Jacobs' results. If so, Michels' PK is quite susceptible to suggestion. These incidental findings about Michels' ability are a v. aluable addition to knowledge about this PK star. The ATS method will be explored further. EVENT-RELATED BRAIN POTENTIALS AND A PHENOMENOLOGICAL MODEL OF PSI-CONDUCIVE STATES Norman S. Don,t Charles A. Warren,t Bruce E. McDonough, and Thomas F. Collura (Kairos Foundation and the University of ll'linois at Chicago). "'Address correspondence to Norman S. Don, Ph.D., 405 N. Wabash Avenue, Suite 4803, Chicago, IL 60611.) A high-scoring subject, Olaf Jonsson, identified three altered states of consciousness associated with psi. He termed these Condition One, Condition Two, and Condition Three, the last being the prime state of psi. Through the process of participant- observation a description of these states was elaborated which was similar to Carl Jung's writings on the psychological aspects of medieval alchemy. Two experiments were performed to investigate the event-related brain potentials (ERPs) associated with this phenomenological model and an objective, clairvoyant task. The term ERP is here defined broadly as that transient brain electrical activity that occurs time locked to specific events--either related to external events, endogenous events, or overt motor behavior. The purposes of this single-subject study were: first, to record brain-wave data from several increasingly psi-conducive al- C? tered states of consciousness (State Study), and a forced-choice, five-response clairvoyant task (Prediction Study); secondly, to perform analyses in the time and frequency domains to discriminate 00 correct (RTE) from incorrect (WRO) guessesa) and the various states of consciousness; thirdly, to cross-validate predictive models on additional, independent RTE and WRO trials. Elaboration of the altered states description was derived from p articip ant -observation of a subject C14 who had performed at above- chance levels in previous psi tasks (Warrena) and Don, RIP 1986, 56-61). The report by Warren and Don presented evidence that both time- and frequency- domain ERP data could be used to sig- nificantly discriminate RTE from WRO trials in a clairvoyant task. a) The present study continued the investigation0. of ERP effects and in addition also studied the subject's Q self-reported altered states of consciousness and the ERP relationship between altered states and clairvoyant performance. In the first experiment (State Study), ERPs were recorded LC) while the subject cycled repeatedly through these three states and a control condition (mental arithmetic). In the Prediction Study 00 ERPs were recorded during a clairvoyant task. In both experi- ments ERPs were recorded over an 8000 msec interval prior to each subject response from 5 EEG scalp sites, one bipolar EOG montage C14 (below and above the right eye), and one EIVIG montage. Electrodes were placed at Fz, C3, Pz, C4. and Oz to scalp sites referred to linked ears using a forehead ground. Grass model 7P122 amplifiers with eight-second time constants were used in a Grass Model 78 poly- graph. On the EMG channel only, the time constant was 0.2 sec.. The one-half amplitude high frequency cut-off was set at 60 Hz on 0 the EEG and EOG channels and 75 Hz on LL the EMG channel. A 60 Hz notch filter was used on the EEG amplifiers. The EEG, EOG, and EMG data and event channel were digitized at 125 samples/sec > (8 msec per datum). A trial consisted 2 of 8000 msec of data (1000 data points) per channel. CL CL In the State Study the subject initiated the beginning of a new trial each time he entered a new Condition (in his subjective