9EHAVIORA L N printed in the Unite WPOY4[Mg hVdIC~10'9'-.-t'?A-RDP96-00789ROO2200110001-5 d St.*PIPYR%c The anomaly called psi: Recent research and criticism K. Ramakrishna Rao and John Palmer Institute for Parapsychology, Box 6847, College Station, Durham, N. C. 27708 Abstract: Over the past hundred years, a number ofscientific investigators claim to have adduced experimental evidence for "psi" phenomena - that is, the apparent ability to receive information shielded from the senses (ESP) and to influence systems outside the sphere of motor activity (PK). A report ofone series ofhighly significant psi experiments and the objections ofcritics are discussed in some depth. It is concluded that the possibility of sensory cues, machine bias, cheating by subjects, and experimenter error or incompetence cannot reasonably account for the significant results. In addition, less detailed reviews of the experimental results in several broad areas ofpsi research indicate that psi results are statistically replicable and that significant patterns exist across a large body of experimental data. For example, a wide range of research seems to converge on the idea that, because ESP "information" seems to behave like a weak signal that has to compete for the information-processing resources of the organism, a reduction of ongoing sensorimotor activity may facilitate ESP detection. Such a meaningful convergence of results suggests that psi phenomena -ay represent a unitary, coherent process whose nature and compatibility with current physical theory have yet to be determined. The theoretical implications and potential practical applications of psi could be significant, irrespective of the small magnitude of psi effects in laboratory settings. Keywords: eWrvoyance; extrasensory perception (ESP); methodology; parapsychology; psi, psychokinesis (PK); replication; scien- tific method; telepathy 1. Introduction There is a large and growing body of experimental liter- ature devoted to the study of certain anomalous interac- tions that seem to involve psychologically meaningful exchanges of information between living organisms and their environment. We call these interactions anomalous because they appear to exceed somehow the capacities of the sensory and motor systems as these are presently understood. These interactions are collectively desig- nated by the term psi. Parapsychology is that branch of science that makes a systematic study of psi anomalies. In other words, it is the business of parapsychology to find explanations of psi anomalies. through scientific inquiry. Psi is traditionally divided into vmious subcategories, each of which has been the subject of experimental research. For example, parapsychologists have been test- ing whether subjects can acquire information that is shielded from their senses (extrasensory perception, or ESP) and whether subjects can directly influence exter- nal systems that are outside the sphere of their motor activity (psychokinesis, or PK). Experimenters have also sought to differentiate forms ofESP, such as "telepathy" (ESP for another's thoughts) and "chdrvoyance" (ESP for external objects and events). ESP is sometimes reported to be time-displaced, in that the information may relate to a past event ("retrocognition") or a future event ("precog- nition"). In practice, it has often proved difficult to isolate these forms of psi experimentally, and nowadays they tend to be defined operationally rather than theoretically (e.g., it is clairvoyance when you. do not have someone c'69efitiffig' the target).. Somewhat contraq to common usage, we are not using the term psi to imply that the anomalous interactions are necessarily "paranormal," but rather that no adequate conventional explanation of the interactions has yet been offered. Phrases stating or implying the "existence" Of psi will be used somewhat informally to indicate that certain interactions have achieved this status. The term paranormal has been a source of some confusion both within and outside parapsychology, and thus we feel that a few comments on the term are in order. Paranormal was first discussed in relation to psi by the philosopher C. D. Broad (1953; 1962; see also Braude 1979b), who defined psychical research (the earlier term for parapsychology) as "the scientific investigation of ostensibly paranormal phenomena" (Broad 1962, p. 3). Broad was careful to use the term "ostensibly paranor- mal," by which he meant phenomena that seem prima facie to conflict with one or more ofwhat he referred to as the "basic limiting principles" of nature. These are not the same as the laws of nature, but father a more funda- mental set ofassumptions that "we unhesitatingly take for granted as the framework within which all our practical activities and our scientific theories are confined" (Broad 1953, p. 7). For example, the assumption that "it is impossible for a person to perceive a physical event or a material thing except by means of sensations which that event or thing produces in his mind" (Broad 1953, p. 10) is a basic limiting principle that govems our way of acquir- ing knowledge. A case of ESP, therefore, would be ostensibly paranormal; it would be genuinely paranormal only when and if it could be shown to really conflict with one or more of the basic limiting principles. It is the task ,proved Fp4r -RDP96-00789ROO2200110001,5§9 .559119A~#gO00/08/08: CIA Ap -0 1987 Cambridge Unive 1 ress 0 nao cc raimer: Varapsycholopy review Approved For Release 2000/08/08 of parapsychology, according to Broad, 1. to investigate ostensibly paranormal phenomena, with a view to dis- covering whether they are or are not genuinely paranor- mal phenomena" (Broad 1962, p. 5). Although Broad's reasoning is sound, the term para- normal has led to some difficulties in practice. For exam- ple, as noted above, it is commonplace to find the terms psi and paranorrial phenomena being used interchange- ably, implying that parapsychology has no subject matter unless the paranormality of the phenomena is accepted in advance! A second and subtler difficulty is more directly related to the term itself. By stressing the conflict be- tween potential "paranormal" explanations of psi and .. normal" science, and at the same time failing to acknowl- edge that what constitutes normal science is historically relative (i.e., it can change from one historical period to the next), the term paranortnal leaves the connotation that explanations that violate the basic limiting principles are unscientific in some fundamental sense. This, of course, is not true. If a "paranormal" theory of psi were someday to be confirmed, the practical consequence would be a redefinition of "normal" science to accommo- date the new theory. In other words, the "paranormal" would become "normal," and the distinction would break down. A similar objection to the term has recently been raised by Paul Kurtz (1981), a well-known critic of parapsychology. It is our view that potential explanations of psi that violate the basic limiting principles of nature are scien- tifically legitimate and, along with conventional explana- tions, should be entertained from the outset in our efforts to explain psi anomalies. Such explanations, unorthodox as they may be, are nonetheless worthy of consideration for the simple reason that psi anomalies seem to violate the basic limiting principles prima facie. Things are not always what they seem, but the possibility that they are should certainly be considered. Thus, the distinction to which paranormal refers is a valid one, even though the term itself is problematic. Recently, Palmer (1986b) has proposed a neutral term, omega, to idento potential explanations of psi that go beyond the basic limiting principles. Thus, "paranormal" explanations would be labeled "omegic." Despite our reluctance to introduce neologisms, we think in this case an exception may be justified. 2. Background Like conventional psychology, experimental parapsy- chology grew out of a need to account for people's experiences in the "real world. " The first major survey of such experiences was conducted under the auspices of the British Society for Psychical Research in the last century (Gurney et al. 1886/1970). More recently, a survey conducted by the National Opinion Research Center of the University of Chicago revealed that a majority of Americans thought they had experienced one or more psychic events in their lives (Greeley & Mc- Cready 1975), Similar results have been obtained in other surveys in the United States (e.g., Palmer 1979), Europe (e.g., Green 1960; Sannwald 1963; Haraldsson et al. 1977), and Asia (e.g., Prasad &Stevenson 1968). Palmer's survey further revealed that for many of those who CIA-RDP96-00789ROO2200110001-5 reported psychic experiences, these significantly influ- enced their feelings, attitudes, and decisions in other areas of their lives. Whatever the explanation of psychic experiences, they happen, they are common, and they are often important to people. For these reasons alone, they deserve serious attention from scientists involved in the study of human behavior and cognition. Although some parapsychological research has di- rectly examined the evidential value and characteriza- tion of these spontaneous psychic experiences (e.g., Hart 1954; Rhine, L. E. 1962; Schouten 1982), the bulk of the research has been experimental, and we will limit ourselves to the latter in this target article. The first major experimental investigation of psi was conducted at Stanford University by John Coover (1917), Sustained research, however, did not begin until 1927, when J. B, Rhine arrived at Duke University to work with William McDougall. With the publication of J. B. Rhine's (1934/1973) monograph Extrasensory Perception, a sci- enfific claim for the existence of ESP was made. It gave the field "a shared language, methods, and problems" (MeVaugh & Mauskopf 1976), and it provided "radical innovation and a high potential for elaboration" (Allison 1973, p. 39). Rhine's procedure was to have subjects guess the randomized order of the cards in a deck containing five examples of each of five geometric symbols: a star, circle, cross, square, and wavy lines. By chance, the subject should get 5 correct out of the 25. Standard statistical techniques were used to determine the likelihood that any given number of hits was statistically sighificant. If the average number of correct guesses per run of 25 exceeded 5 to a significant degree, and acquisition of information by artifactual means such as sensory cueing and logical inference was ruled out, ESP was considered to have been demonstrated. Using this methodology, Rhine (1934/1973) reported highly significant results, especially with five selected subjects who were tested repeatedly over a number of years. Prior to August 1, 1933, all subjects in the program had completed a grand total of 85,724 trials, with an average score of 7.1 hits per run. . I The reaction of the scientific community to Rhine's claim was understandably cautious and critical. Subse- quent to the publication of the monograph, there were 35 criticisms contained in 56 published reports. Some of these criticisms were specific and others were merely speculative. The specific criticisms had to do with Rhine's methods of data collection and statistical analysis. These criticisms and Rhine's responses are fully documented in the book Extrasensory Perception After Sixty Years (Rhine et al. 1930). Thc-flot line of criticism dealt with the experimental conditions. One essenti requirement or-an ~c~e-ptabie ES7 -experiment was that data should be collected under conditions that provide no reasonable opportunity for sensory leakage of information or inferential knowledge of the targets. Skinner (1937), Wolfle (1938), and J. L. Kennedy (1938), among others, pointed out that under certain lighting conditions the commercially produced ESP cards could be read through their reverse sides. Rhine responded that the original experiments were conducted with hand-printed ESP cards that were free from such defects and that in his more formal experiments 540 Upy(g)(LeAq JFE39T1J3~JpA*R94PP.Q/08/08 : CIA-RDP96-00789ROO2200110001-5 Liao & Miner: Parapsychology review l/08/GbudCIASRbP"uN'MOM&22&0n"g"jt a " ' the use of screeYAJ# Ic r1e rh i b1 t t l bt i t Ii ti b " i U f 3) th j a - e men n a e a car S e ra ng a s e o n o e exper mn ,, Y o fro respec ca rep on; ( (1438), Kellogg (1936), servations in and Leuba (1938) parapsychology argued that an are not unrelated, and sig- increase in the experimentalor of ESP researchnificant patterns rig had involving large bodies of experimental resulted in a correspondingine in ESP results,data are apparent; decl sug- (4) a wide range of process-oriented gesting that extrachance research has nitive ESP scores were due focused on process to loose a single cog that experimental conditions. may be seen even To this Rhine responded to give coherence a that and degree of his most rigorously consistency controlled experiment, to a diverse the Pearce- array of experimental results, and Pratt series, did (5) the small give highly significant magnitude of results (Rhine et most current psi effects is al. 1940). Although irrelevant to this experiment was both their later challenged theoretical importance and their by critic C. E. M. potential applicability. Hansel (1966) - with questionable success (Hansel 1980; Rhine & Pratt 1961; Stevenson 1967) - as being susceptible to fraud on the part of the subject, it was still 3. The question more rigorously controlled of the "conclusive" than the experiment other experiments in the original data base and thus supported Rhine's Referring to point. parapsychology, * Phillip H. I- Th Abelson (1978), t Edit d li f S f i d t i d U l S i N iti d i d t Id l e secon or o ne o c cr ence, a . a ana . c ews an sm re in e Wor o s quote a s.. ys Wif1o`u-g_Rbk__~19_35),_Ke1logg Report as saying (1936), Heinlein that "extraordinary and claims require ex- Heinlein (1938), Herr traordinary (1938), and Lemmon evidence." (1939) crit- This statement implies that the icized various features strength of of the statistical evidence required analysis used by to establish a new phe- Rhine and his colleagues. nomenon is directly In particular, the proportional criticism to how incongruent the focused on Rhine's phenomenon is assumption that the with our prior binomial theorem notions. Our prior notions, is applicable to "closed however, are decks," decks in not always which the self-evident truisms. They are number of times each derived from, type of card appears among other is not free to things, prevailing religious vary. This aspect and cultural of the methodological beliefs, personal debate essentially experiences and observa- ceased in 1937, when tions, and our Burton Camp, President general world of the view. They are translated Iustitute of Mathematical into subjective Statistics, stated probability that Rhine's estimates and determine the 11 statistical analysis evidential demands is essentially valid. we make for if the Rhine inves- a given claim. If the tigation is to be subjective probability fairly attacked it of a disputed must be on other claim is zero, than then no mathematical grounds" amount of empirical (Camp 1937). For evidence will further details, be sufficient to estab- see Burdick and Kelly lish that claim. (1977). In serious scientific discourse, however, It would be wrong few would be to conclude from expected to this, however, that take a zero-probability stance Rhine's experiments because such were perfect and a stance could that they had be seen to be sheer dog- conclusively eliminated matism and the every alternative very antithesis explanation. In of the basic assumption of retrospect, one could science's open-endedness. suggest improvements in the ex- perimental conditions Nevertheless, of his experiments. the demand But for his for extraordinary evidence time, Rhine's best of psi often experiments were seems to be ahead of others in derived from an implicit notion of the behavioral sciences. its a priori y. For The experimental impossibilit example, precautions he some critics of psi took, including two-experimenter research have ed a "foolproof' controls and double- demand experiment that blind procedures, would control were rare in other for all conceivable disciplines at that kinds of error, including time. Nonetheless, fraud by the much of the early experimenter(s). criticism of Rhine's They have argued that if a experiments was helpful claim is made in progressively for the existence raising the stan- of a phenomenon that dards of ESP research conflicts with and reducing the "established possibility of laws," it is much more par- experimental errors simonious to and artifacts. assume error or even fraud on the part of the Since the publication claimant than of Rhine's monograph it is to assume over fifty the reality of that phe- years ago, there have nomenon (Price been hundreds of 1955; Hansel experimental 1966). This argument is reports of evidence often identified for psi. Yet skepticism with David has not de- Hume's (1825) maxim that creased. Psi results .1 no testimony are generally ignored is sufficient in,mainstream to establish a miracle, unless science, and when the testimony called to the attention be of such of scientists they a kind, that its falsehood would are apt to arouse be more miraculous suspicion. When specific than the fact criticisms are which it endeavours to voiced, they generally establish" (p. include the following: 115). Hume's (1) There is maxim is a metaphysical no "conclusive" experiment statement, and in parapsychology's it is inappropriate long to use it when one history; (2) there speaks of empirical is no repeatable evidence. Moreover, psi experiment; (3) his definition the of so-called significant a miracle as psi results are disparate, a universally incoherent, nonexistent event is self-contra- and isolated one-shot dictory inasmuch pport observations that as any claimed of do not merit evidence in a su scientific attention; miracle is also ity (4) the results themselves evidence against of are nonsen- the universal its sical in that they nonexistence do not suggest any (Rao 1981a). lawful relationships As Saint Augustine or remarked, progressive research "Miracles occur programs; and (5) in contradiction even if psi is real, not to nature, but to it is too weak to what is known be of any practical to us of nature." importance. If such It should also be kept in perceptions were strongly mind that Hume supported by all might not have the available regarded psi phenomena data, it would be as miraculous right to ignore parapsychology's or as anything claims. more than extraordinary But the fact (as we events. hope to show in the following pages) is that (1) there are The call for good experiments a totally "foolproof' that seem to provide study assumes that at a evidence for the existence given time one of psi by reasonable can identify standards all possible sources of error Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2200110001-5 BEHAVIORAL AND BRAIN SCIENCES (1987) 10:4 541 .MR%pg)JffftrrAWase 2000/08/08 CIA-RDP96-00789ROO2200110001-5 Rao & palme#, afid how to control against them. Such a methodological stance is comparable to the epistemological position that one can determine for all time to come what is and is not possible. Again, the demand for experimental controls against experimenter fraud is unique to discussions of evidence for what are perceived to be extraordinary claims. Pushed to its extreme, the hypothesis of experi- menter fraud becomes nonfalsifiable, in that it is impossi- ble to be certain that fraud is completely eliminated in any given experiment. The concept of a "conclusive" experiment, totally free of any possible error or fraud and immune to all skeptical doubt, is a practical impossibility for empirical phe- nomena. In reality, evidence in science-is-,A, matter of degree; t e c t t one can concoct a ternative exp - ana- tid-n-sof-a--finding does'n-kot ati-tomati6illy- r-e-rider -that finding evi*dJent ally _~W~fthless. Evidentiality must be -as- sessed on a continuum and in relation to the plausibility of and the empirical support for the competing hypotheses. These considerations demand that a "conclusive" experi- ment be defined more modestly as one in which it is highly improbable that the result is artifactual. In this sense, we think a case can be made for "conclusive" experiments in parapsychology. 3.1. Schmidt's REG experiments A defense of the existence of probabilistically conclusive parapsychological studies requires a detailed review and discussion of any experiments that might qualify. Because such a treatment must be rather lengthy, we will limit ourselves to a single group of experiments as an example. Although they are somewhat dated, we have chosen Helmut Schmidt's (1969a, 1969b) reports on random event generator (REG) experiments because (a) they represent one of the major experimental paradigms in contemporary parapsychology; (b) they are regarded by most parapsychologists as providing good evidence for psi; and (c) they have been subjected to detailed scrutiny by critics. In no sense do we imply that these are the only good experiments the field has to offer. Nor do we believe, for the reasons stated above, that there can be any crucial experiment or experimental program on which the case for psi does or could rest exclusively. At the time of conducting these experiments, Helmut Schmidt was a physicist at Boeing Scientific Research Laboratories. The studies were designed to test the possibility of ESP and were carried out with the help of a specialty built machine that seemed to rule out all ar- tifacts arising from recording errors, sensory cues, and subject cheating. The safety features of the Schmidt machine are actually superior to those of the VERITAC machine used earlier by Smith and his colleagues to test for ESP (Smith et at. 1963). Hansel (1966) had praised VERITAC as "admirably designed" and had suggested that it -could be "standardized for testing subjects for extrasensory perception" (p. 172). The Schmidt machine randomly selected targets with equal probability and recorded both the target selections and the subject's responses. The subject's task was to guess which of four lamps would light and to press the corresponding button if he was aiming for high scores (or to avoid that button if aiming for low scores). As Schmidt (1969b) described it: Approved For Release 2000/08/08 During a test, the subject sits in front of a small panel with four pushbuttons and four corresponding colored lamps. Each of the pushbuttons simultaneously acti- vates a recorder switch and a trigger switch. The recorder switch serves to register which of the buttons has been pressed. The four trigger switches are con- nected in parallel such that pressing any one of the buttons closes a circuit, in turn triggering the random lighting of one of the four lamps. The system is de- signed so that on repeated pressing of the buttons the lamps light in random sequence, i.e., each lamp lights with the same average frequency, and there is no correlation between successively lit lamps or between the buttons pushed and the lamps lit. (p. 101) Random lighting of the lamps was achieved, following the subject's response, by a sophisticated electronic ran- dom event generator that used a radioactive source, strontium 90. (See Schmidt (1970b] for a more complete account of the hardware design and methods of statistical evaluation.) The REG was extensively tested in control trials and found not to deviate significantly from chance. The sequence of buttons pressed and lamps lit is recorded automatically on paper punch tape. In the research reported here, the two types of test (trying for a high or low number of bits) were recorded in different codes, such that the evaluating computer could dis- tinguish between them. The number of trials made and hits obtained were displayed to the subject by elec- tromechanical reset-counters. These numbers were also registered by nonreset counters, and the readings of all counters were regularly recorded by hand. This record agreed with the results obtained from the paper tape. The equipment was fraud proof, so that one could, in principle, let the subjects work alone. This was done, however, only in a small part ofthe tests with subject OC in the first experiment and did not increase the scores. In all other tests the writer was present in the same room with the subject. (Schmidt 1969b, p. 103) Schmidt's first report was based on two experiments. The subjects in this study were preselected on the basis of their performance in the preliminary tests. In the first experiment there were three subjects. All of them at- tempted to obtain high scores. Together they did 63,066 trials and scored 16,458 hits, which was 691.5 more than mean chance expectation (MCE). The probability that such a result occurred by chance is smaller than 2 X 10-9. In the second experiment, two subjects from the first series and one new subject participated. One subject aimed for high scores and another for low scores. The third aimed high in some trials and low in others. The total number of trials was 20,000. Of these, 10,672 were high-aim trials and 9,328 were low-aim. The combined deviation of hits in the desired direction was 401 greater than MCE, which has an associated probability smaller than 10- 10. In the third experiment, Schmidt (1969a) tested six subjects, including two who had participated in the trials just described. The experiment was designed to test primarily for clairvoyance; the targets were digits from a random number table further shuffled by a congruential generator and recorded on paper punch tape. The sub- jects com leted a total of 7,091 hi h-aim trials and 7,909 : CIAMP96-00789RO02'1001 10001 -5 5 4 21 9F",AV!('PAL -N:C, BR41', ~C~E~:CEc_z ~1987~ `0 4 Rao & Palmer: Para~s~chologry review A d F R l 2000/08/08 CIA RDP96 00789F 00110001 5 or - e - ease - pprove w-Win trials, for a grand er could total of 15,000. The combined have artifactually produced the significant re- -viation of hits in the SuIts. desired direction was +260 (p 3 X 10-6). Hansel's criticism (2) of the machine itself overlaps criticism (1-b) above and was discussed under that heading. 2. Criticisms of Schmidt's The final REG experiments reason given by Hansel for his rejection of Schmidt's results was that they have not been confirmed. ansel (1980) discussed But this the "weaknesses" in Schmidt's again seems erroneous, as will be shown in (periments under three Section 4. headings: (1) experimental 1. 1 below. de- Hansel made no mention of several gn, (2) unsatisfactory experimental features of the machine, reports and (3) already in the literature that did in iability to confirm the fact claim findings. He criticized to confirm the experi- Schmidt's results; he instead re- iental design (a) for its ferred only failure to specify in to the 1963 advance the report of Smith et al., which gave exact numbers and types null results of trials to be undertaken when VERITAC by was used to test for ESP. But ach subject," (b) for its even this introduction of high-aim comparison and low- is problematic. First, the machines, im conditions, and (c) experimental for its lack of control procedures, of the and manipulation of the psy- xperimenter. chological conditions differed markedly between the two Strictly speaking, criticism studies. (a) is not relevant to Second, the Schmidt's subjects were carefully .iain purpose of the experiment, screened which was to determine through pretesting procedures, whereas those tot whether a given subject who participated had ESP, but whether the in the VERITAC experiment were not. !xperiment as a whole provided In a more evidence for ESP. It is recent publication, Hansel (1981) proposed a rue, however, that in Schmidt's scenario first experiment the that permits the possibility of trickery without iumber of total trials providing was also not specified any evidence precisely in that fraud had indeed occurred. idvance. The high level nificanceReferring of statistical sig ob- to one of Schmidt's -experiments testing PK ained, however, renders hat this (Schmidt the possibility t factor 1970a), he claimed that the subject could have -ould account for the results shorted "either extremely unlikely. And, the + 1 as or the -I input in the display Hansel acknowledges, this panel to problem was corrected the earth in the line according to whether he wished to ater experiments. produce a high or a low score" (p. 30), which would Criticism (b) is not substantiated. account for Noting that high-aim the significant results. This argument seems icores gave a positive fallacious. deviation and low-aim Because scores a the REG and electronic counters negative deviation, Hansel were sealed argued, "The fact that in a metal when box and the REG outputs were positive and negative deviations completely are combined (maintain- buffered, there was no way the subject could ing their sign) they invariably have tampered give a purely chance score with the apparatus in the way Hansel suggests that sampling suggests. from a common distribution Second, may the data were independently recorded have taken place" (p. 230). on punch pe In the first place, this tape. Had ma- argu- the subject shorted the ta ment fails to account for chine, the iffered Experiment 1, which involved total number of punches would have d only the high-aim condition from the and gave results that 128 specified were for each run. Inspection of the just as significant as tapes revealed in the other experiments. no such Second, it discrepancies (Schmidt, personal is not clear how Hansel's communication). criticism could apply to the other experiments, since Hansel went the high and low conditions on to argue that the experimenter himself were assigned in advance could have y affected the and recorded automatically easil punched record. on This is paper punch tape in different debatable, he possibility codes. it would seem, but t that Schmidt in could have fact, that the introduction faked his of high/low conditions data somehow has a has already been acknowledged. certain additional merit Recently, in that one condition however, could be Schmidt has published a PK experi- considered as a control ment designed for the other, as well to rule as for out the possibility of his (or his two machine bias. It is of co-experimenters) interest that in discussing falsifying a different the data without collabora- Schmidt experiment, Hansel tion from (1981) himself criticized at least one of the others (Schmidt et al. 1986). Schmidt for not having Briefly, a control condition and Schmidt, recom- located at his lab in San Antonio, Texas, mended the introduction prepared of a condition in which lists of "the paired six-digit random numbers, called subject would not be 'willing' seed numbers, the light to move, or which were he to be used to generate se- would aim at moving the quences of y light in the opposite quasirandoin means direction" binary digits of b a (p. 32, our italics). complex mathematical y algorithm to known onl Schmidt. Hansel went on to contend These seed that two dfferent ma- numbers were mailed to the private address chines, one for high aim of Professor and the other for low Luther Rudolph aim, should (L. R.) of Syracuse Univer- have been used. But would sity. Robert not such a procedure have Morris (R. M.) of the same university inde- been criticized on the pendently grounds that any obtained obtained dif- a list of random target directions ference between the scores (high and could have been due to low), one the for each binary sequence, by using his opposite bias of the two laboratory's machines? own REG. R. M. and L. R. exchanged their Criticism (c) is valid copies of if by "control of the the target-direction experimenter" sequences and the seed Hansel meant control against numbers and experimenter fraud. It then made the former available to Schmidt. would have been entirely For the test possible for Schmidt to proper, fake the the subject in San Antonio entered results if he had wished the seed to. In the extreme case, numbers for into a computer. The computer then example, the whole experimental derived the report could simply binary sequences, which in turn governed have been fabricated. We the display cannot conceive, however, on a computer screen of a pendulum swinging how a nonintentional rrar-on-th- A PIP U0VW tr Pb+*Ww-6,2M o o 8TOb rardfw-m# *- A 7 89*60 P2bm ivet 14,5 BEHAVIORAL AND BRAIN SCIENCES (1987) 10:4 543 Rao-& PakPP&W0*ReWa'9e 2000/08/08 : CIA.RDP96-00789ROO2200110001-5 pendulum to swing with large amplitude on high-aim trials and with small amplitude on low-aim trials. I At the end of the run, which lasted for about a minute, the display showed the average swing over the run; thus the subject was given feedback about his rate of success. Schmidt et al. reported significant results in support of their hypothesis. The combined Z for all the ten sessions was 2.71 (p < .005). Because (a) the seed numbers for the binary sequences and (b) the target directions were inde- pendently derived by Schmidt and Morris, respectively, we know of no way Schmidt or Morris alone could have artifactually obtained the results. Such security pro- cedures involving experimenters working independently in two different laboratories are seldom used in scientific research; but it is understandable that Schmidt felt that the validity of his results should not be based ultimately on his honesty alone. Of course, the possibility of fraud is still not eliminated completely in this experiment. Even if we grant that Schmidt alone could not have faked the results, it remains possible, though less probable, that Schmidt and Morris, or Morris and Rudolph, could have conspired to produce them spuriously. Perhaps the logical next step is to have a critic participate as a co-experimenter, using the design of Schmidt et al. We would be curious to see how critics would react if such an experiment succeeded. Hansel's criticisms of Schmidt's experiments are rou- tinely taken as valid by most writers skeptical of psi (e.g., Alcock 1981). One of the few critics of psi who questions the basic premises of Hansel's reasoning on this point is Hyman (1981). "There is no such thing as an experiment immune from trickery," says Hyman. "Even if one as- sembles all the world's magicians and scientists and puts them to the task of designing a fraud-proof experiment, it cannot be done" (p. 39). Hyman, however, agrees with Hansel that Schmidt's PK experiments "do not provide an adequate case for the existence of psi" (p. 34). His principal reasons are twofold: (1) "Experience shows that the most promising research programs in parapsychology will most likely be pass6 within a generation or two" (p. 37); and (2) although Schmidt's randomization tests con- trol against "long-term, or even temporary" machine bias, they do not "control against possible short-run biases in the generator output" (p. 38). He suggested, as did Hansel, that matched experimental and control se- quences would have been a superior procedure. The first point is not really a substantive criticism but merely counsels patience. The same thing can be said of research in some other areas of psychology. Moreover, 1. passd" does not necessarily mean "discredited," and much of the older research in parapsychology has with- stood criticism rather well. The second point, as Hyman himself recognizes, "does not automatically provide an alternative explanation for how Schmidt obtained his results" (p, 38). Schmidt, who was aware of such a possibility, notes that "many more randomness tests were done than published to satisfy my own questions about the possibility of temporary random generator malfunctions" (Schmidt 1981, p. 41). Also, it is difficult to see how such malfunctions could account for subjects' ability to anticipate the timing and direction of the hypothesized short-run biases in Schmidt's early PK research, which used a high C%H (Schmidt 19APPrOdy-, _iFqdTbA00WtW1(mo1re108 recent work, direct comparisons were made between experimental and control sequences (e. g., Schmidt 1976). 4. The question of replication Even assuming that it was possible to determine con- clusively the proper interpretation of a single experimen- tal result, such an exercise would have little value in the context of doing science. The way the scientist functions is different from the way the historian does, for example. Unique events and isolated facts, unless they lead to, or are capable of leading to, some kind of general law, ordinarily hold little interest for science. Unlike historical facts, most phenomena of science are capable of being repeated. The Battle of Gettysburg will not be fought again. But psi as a laboratory effect must be reasonably capable of being observed repeatedly if one is to study it effectively and to understand it. Thus, as even Hansel (1980) concedes at one point, the importance of a fool- proof experiment recedes into the background as the phenomena become increasingly replicable. Replicability does not necessarily mean that a finding must be reproducible on demand. It is not strictly an either-or situation, but a continuum (Rao 1981b). In this sense of statistical replication, an experiment or an effect may be considered replicated if a series of replication attempts provides statistically significant evidence for the original effect when analyzed as a series. It may be argued that statistical replication is simply imperfect replication, and that a real phenomenon is something that is in principle repeatable. If a phe- nomenon has occurred once, it will occur again, provided the same set of circumstances arises. If one had perfect understanding of the critical variables, one could invari- ably predict its occurrence; if one had control over those variables one could produce the phenomenon on de- mand. The problem is that, in practice, perfect duplica- tion of conditions is impossible to achieve. This is es- pecially true in behavioral science experiments, where the causes of an effect are likely to be complex and difficult to pin down. This does not mean that replicability cannot be im- proved' substantially if some understanding of these cru- cial variables can be achieved. Indeed, such understand- ing is a major goal of scientific investigation. The other side of the coin, however, is that inquiry in such cases begins without this understanding. It is therefore inap- propriate to demand absolute or even strong replicability of a phenomenon simply as a prerequisite for 1) research. further 4.1. Examples of replicabliffy In parapsychology Once we *ve u the notion of absolu - replication, we can see that parapsychological phenomena are replicated. in a significant statistical sense. For example, Palmer's (1971) review of so-called sheep-goat studies reveals that in 13 of the 17 experiments that used standard methods of analysis, the "sheep" (the subjects who believed in the possibility of ESP) obtained higher scores than did the 1. 1. 0R*WQG"-,5Ath 6 P*40ft9we Ae%4'ac lieving statistical significance. Carl Sargent's Approved For Release 2000/08/08 p8i) review of the reports published in English on the association between ESP and extraversion suggests that significant confirmations of a positive relationship occur at over six times the chance rate, However, the most extensive evidence for the statistical replicability of psi comes from the three data bases to be discussed in more below. detail 4.1-1. REGs and psi. Since the publication of the REG results discussed in Section 3.1 above, Schmidt has car- ried out several other successful REG experiments, mostly involving PK. More to the point, a number of other experimenters have successfully used the same devices or similar ones to test for psi. The most prominent of these replications comes from the laboratory of Robert Jahn at Princeton University (Jahn 1982; Nelson et a]. 1984). Jahn and colleagues use an REG based on a commercial electronic noise source. The hits are counted and displayed on the instrument panel and are permanently recorded on a strip printer as well as a computer. The subject's task is to influence the device mentally to produce an excess of hits on predesig- nated PK+ trials and an excess of misses on PK- trials. In a total of 195, 100 PK + trials, 22 subjects obtained a mean score of 100.043 (MCE = 100). The mean for the same number of PK- trials was 99.965. Although small in magnitude, both these means are significantly different from mean chance expectation. The combined proba- bility of the results is approximately 3 X 10-4. Each trial in Jahn's experiments incorporated alternate positive and negative counting on successive samples to provide an on-line internal control against any systematic bias in the noise source (i.e., positive and negative noise pulses alternated as hits). Also, baseline trials were re- corded "under a variety of conditions before, during, and after the active PK trials" (Jahn 1982, p. 148) in a manner resembling that recommended by critics. The mean score for these 179,250 baseline trials was 100.005, which does not differ significantly from chance. Radin et al. (1985) conducted a preliminary survey of all binary (two-choice) REG experiments published from 1969 (the year of Schmidt's first published REG experi- ment) to 1984. The sources sampled were the five major refereed parapsychological journals, the bound Proceed- ings of refereed papers presented at the annual Para- Psychological Association Conventions, and a report of the Princeton data by Nelson et al. (1984), cited above. The reviewers defined an "experiment" as the "largest possible accumulation of data compatible with a single 'direction of effort' assigned to the subjects" (p. 205). In other words, data from all trials in which subjects aimed for the same binary outcome were pooled, ignoring other experimental conditions or classifications that may have pertained. The reviewers uncovered 56 reports from approx- imately 30 principal investigators describing a total of 332 individual experiments. For 30 of the nonsignificant ex- periments, the authors of the reports provided insuffi- cient data to allow the outcome (deviation of the hit total from chance) to be expressed quantitatively. In each of these cases, the reviewers randomly selected a Z-score from normal (null) distribution of a the outcome. Seventy-one Rao & Palmer: P h I 0 6arag6 o o% review : CIA-RDP96-00789R 2 11 01-5 results significant at or beyond the 5% level (2-tafled), and the combined binomial probability for all the studies was 5.4 X 10 - 43. The outcome was still significant, although more modestly so, when the 'data from Schmidt and the Princeton group were removed (p < 4.25 X 10 4.1.2. Ganzfeld and ESP. A second major research para- digm in which the replication rate over a relatively large number of studies has been systematically evaluated concerns ESP in the ganzfeld. The ganzFeld is a homoge- neous visual field produced, for example, by placing a halved Ping-Pong ball over each eye with cotton filling around the edges. While the subject relaxes in a comfort- able chair or bed, a uniform white or red light is focused on his face from about two feet. Sometimes the subject also listens to "pink" noise through attached earphones. Subjects typically report a pleasant sensation of being immersed in a "sea of light", (Honorton 1977, p. 459). In a typical ganzfeld-ESP trial, the subject receives approximately 30 minutes of ganzfeld stimulation. After a period of adjustment and relaxation, the subject is asked to report all images, impressions, and so on, that occur at the time. From another room, an experimenter blind to the target monitors the subject's mentation via a micro- phone link and a one-way mirror. In a room located some distance from the subject, another experimenter acts as the agent. Some time after the subject has been in the ganzfeld, the agent-experimenter opens an envelope containing a target picture (randomly chosen from a pool of four), views it for about 15 minutes, and then stays in the room for an additional 10 minutes. After the comple- tion of the ganzfeld period, the first experimenter gives the subject four pictures and asks him to assign them ranks of 1 through 4 for their correspondence to his mentation. At this time neither the subject nor the first experimenter knows which of the four pictures is the target. The agent-experimenter is then called in and reveals the target picture. The first ganzfeld experiment in parapsychology was reported by Honorton and Harper (1974). The results of this experiment were subsequently replicated by Terry and Honorton (1976), Braud et al. (1975), and Sargent (1980), among others. According to a recent count adopted both by Honorton (1985) and critic Ray Hyman (1985b), there are 42 published ESP experiments that have used the ganzfeld procedure. After correcting for multiple analyses, if any, Honorton concluded that 19 of the experiments (45%) gave significant evidence for psi at or beyond the 5% level. Moreover, 26 of the 36 studies for which the direction of the effect could be clearly deter- mined (72%) gave deviations, in the positive direction, as compared to the 50% expected by chance. Hyman (1985b) dissented, concluding that the "rate of 'suc- cessful' replication is probably very close to what should be expected by chance given the various options for multiple testing exhibited in the data base" (p. 25). Later, however, he came to agree with Honorton that "there is an overall significant effect in this data base which cannot reasonably be explained by selective reporting or multi- ple analysis" (Hyman & Honorton 1986). Z-scores to represent 4.1.3. The differential effect Another area of psi research with a large number of studies spanning a long period of BEHAVIORAL AND BRAIN SCIENCES (1987) 10:4 545 Rao & Palmer: Parapsychology review A r% e For F sFleaspd1Wrq2J19#J08 ToviO uFJec s o score the tendency MIR s in successive ESP tests when these consist of two con- trasting conditions, such as two different sets of targets or two dfferent modes of response. In other words, subjects score above chance in one condition and below chance in the other. The first authois (K. R. R.'s) initial encounter with dfferential scoring occurred when he attempted to test subjects using both ESP cards and cards consisting of symbols to which the subjects were emotionally attached. In the first experiment, he found not only that the subjects obtained more hits than expected by chance with the cards of their chosen symbols, but also that their scores on cards with ESP symbols were lower than MCE. The scoring pattern with one set of cards was the mirror image of the pattern with the other (Rao 1962). Since then Rao has carried out a large number of tests under a variety of conditions and has found a rather consistent tendency on the part of subjects to show a bimodal response pattern when the ESP test consists of two contrasting conditions (Rao 1965). It is interesting to note that evidence for the dfferential effect can be found in a number of studies carried out before and after Rao's studies, even when the experi- menters themselves were not looking for it. For example, Rao and Krishna (in press) examined 72 independent comparisons between ESP scores obtained by the same subjects responding to two different classes of targets where interactions with other variables had not been predicted. Their sources were the five major refereed parapsychological journals and reports of refereed papers presented at Parapsychological Association conventions. They found that 45 of the 72 comparisons (63%) showed differential scoring, where we would expect 36 (50%) by chance (p < .05). In 19 of the experiments (26%), the scoring rate between the two conditions was significantly different at or beyond the .05 level, though one would expect only 3.6 experiments (5%) to show significant differences by chance. The meaning of the differential effect is not yet clear. It was not derived from a theory or model and provides no explanatory construct that might help us to understand psi. Rather, it reflects a characteristic of psi in a certain type of design, a characteristic that any adequate theory of psi must ultimately account for. One may call it a descriptive construct as distinct from an explanatory construct. Descriptive constructs are important in the early stages of scientific inquiry because, by defining what it is that a theory must explain, they serve to channel the process of theory development. Much of the research in modem parapsychology is directed toward identifying such descriptive constructs or "effects," with the objec- tive of bringing closer to attainment the ultimate goal of a credible theory of psi. 4.1.4. Overview. The proportions of statistically significant studies in the three areas we have reviewed are as follows: REGs (21%); ganzfeld (45%); dfferential effect (26%). Given the expected success rate of 5%, these values are not trivf al, and they compare favorably with comparable examples from psychology, such as the placebo effect (Moerman 1981) and the experimenter expectancy effect (Rosenthal & Rubin 1978). The latter authors, for exam- ple, reviewed evidence on the experimenter expectancy effect in eight types of experiments - The median replica- Approved For Release 2000/08/08 5k topi c (animal learning: 73%), the percentages ranged from 22% to 44%, which is very similar to what we find in para- psychology. 4.2. Some criticisms A number of objections can be raised to the kind of procedure we have used in obtaining these replication rates, objections similar to those that have been raised in discussing experimenter expectancy effects (Barber 1969; 1973). Some of these objections will now be discussed in relation to the data under consideration. 4.2.1. Comparability of studies. One objection to such analyses is that the studies included are often not directly comparable. This objection has merit, but only to a point. We should not insist, for example, that all experiments be strict replications of one another. So long as they con- stitute conceptual replications, methodological differ- ences can often be treated as random variables that actually serve to increase the generality of any conclu- sions that might be drawn from the analysis. On the other hand, it is usually desirable that the outcomes of the studies be represented by, or reduced to, some common metric. One of Hyman's (1985b) criticisms of the ganzfeld data base, for example, was that the studies used diver- gent and sometimes multiple measures of the dependent variable, and that the primary measure was sometimes not specified in advance. In response to this objection, Honorton (1985) computed a new analysis, using as a single, uniform measure Z-scores representing the pro- portion of trials in the experiment in which the subject correctly picked out the target during the judging (i.e., direct hits). This was the measure used in the original ganzfeld experiment by Honorton and Harper (1974), and it was the measure most frequently reported in the data base as a whole. Sufficient information for this analysis was provided for 28 of the 42 experiments in the data base. These experiments came from ten dfferent labora- tories. Twenty-three of the 28 experiments (82%) yielded positive Z-scores, 12 of which were individually signifi- cant at the .05 level on a one-tailed test. The cumulative Z-score for all 28 studies, computed by the Stouffer method (Rosenthal 1984), was 6.60 (p < 10-9). Both Radin et al. (1985) and Rao and Krishna (in press) dealt with the uniformity issue in their analyses of the REG and differential effect experiments (discussed above) by using as a common metric Z-statistics. In the former case, these represented the proportion of trials that were hits; in the latter case, they represented the difference between the proportions of hits in the two conditions. 4.2.2. Publication bias. A second criticism concerns whether these analyses may suffer from biased selection and so-called publication artifact; that is, nonsignificant results may systematically go unreported, and therefore our s~Mtle of studies may not reflect the'irue state of . les 72~1.rs!'Close scrutiny of the field suggests that piu-bTic-a- Tio-n bias cannot explain away the significant number of replications in parapsychology. Parapsychologists are sensitive to the possible impact of unreported negative results, more so than most other scientists. Our profes- CIA-RDP96-00789ROO2200110001-5 546 BEHAVIORAL AND BRAIN SCIENCES (1987) 10 4 Approved For Release 2000/08/08 - CIA-RDP96-00789ROO2200110001-5 Rao & Palmer: Parapsychology review ,ional society, the Parapsychological Association (PA), ias advocated a policy of publishing the results of all nethodologicafly sound experiments, irrespective of out- ~ome. Since 1976, this policy has been reflected in the publications of all the journals affiliated with the PA and in the papers accepted for presentation at the annual PA conventions. This policy, however, cannot guarantee that re- searchers will submit negative findings for publication. Fortunately, thanks to a technique developed by Rosen- thal (1979), we are able to estimate the number of un- published and nonsignificant experiments that would be necessary to reduce an entire data base to nonsignifi- cance. Honorton (1985), for example, used Rosenthal's technique to estimate that 423 nonsignificant ganzfeld studies would be needed to reduce the direct-hit studies in this data base to a nonsignificant level, Given the complex and time-consuming nature of the ganzfeld pro- cedure, it is unreasonable to suppose that so many experi- ments exist in the "file drawer." As noted earlier, Hyman now agrees that selective reporting cannot account for the aggregate findings in the ganzfeld data base (Hyman flonorton 1986). A particularly ingenious way of estimating the extent of the file-drawer problem was implemented by Radin et al. 4985) in their analysis of the REG data base. By inspect- ing a graph of the distribution of outcomes, they noted a marked discontinuity at the Z-value associated with sta- tistical significance: There were too many studies at the tail to make a smooth curve. They determined that the curve could be smoothed by adding 95 nonsignificant experiments to the data base. Doing this reduced the combined binomial probability of all the studies from 5.4 X 10- 43 to 3.9 X 10- 18, still an impressive value. Using the Stouffer method, Radin et al. then estimated that ten parapsychology laboratories would each have needed to produce nonsignificant studies at the rate of 2.6 per month over the 15 years surveyed to cancel out the effect. Finally, there are some areas in parapsychology where we can be reasonably certain we have access to all the experiments done. One such area concerns the rela- tionship between ESP performance and the ratings ob- tained on the Defense Mechanism Test (DMT) devel- oped in Sweden by Ulf Kragh and associates (Kragh & Smith 1970). Because the administration and scoring of this test requires specialized training available to only a few individuals, it has been possible for Dr. Martin Johnson of the University of Utrecht, the leading authori- ty on the DMT and a man very sensitive to the issue of publication bias, to keep track of the number of relevant experiments conducted by qualified persons. In all ten of these studies the less defensive subjects scored higher on the ESP test. In seven of them, this effect was significant at the .05 level, one-tailed Uohnson & Haraldsson 1984). 4.2.3. Controls and flaws. A third line of criticism relates to experimental controls. It is 'argued, for example, that the replication of an experimental result by other experi- menters "does not assure that experimental artifacts were not responsible for the results in the replication as well as in the original experiment" (Alcock 1981, p. 134). It is true, of course, that the replication of an effect implies nothing directly about its cause. But it is also a basic premise of experimental science that repli6ation Approved For Release 2000/08/08 reduces the probability of sonw cauml explanations, par- ticularly those related to the honesty or competence of individual experimenters. As Alcock (1981) himself states in another context, "It is not enough for a researcher to report his observations with regard to a phenomenon; he could be mistaken, or even dishonest. But if other peo- ple, using his methodology, can independently produce the same results, it is much more likely that error and dishonesty are not responsible for them" (p. 133). A more specific set of criticisms has been offered by Hyman (1985b) with reference to the ganzfeld-ESP data base. He concluded that the case for replication in this area is unconvincing because of the presence of meth- odoligical flaws such as potential sensory cues (e.g., including the target handled by the sender in the set given to the subject for judging), suboptimal randomiza- tion of targets (e.g., hand-shuffling), and multiple statis- tical analyses of the data. Honorton (1985) replied that Hyman made several unsupported assumptions in his analysis and interpretation of the ganzfeld-ESP data, and, in particular, that he often did not assign flaws properly with respect to his own criteria, Honorton presented his own analyses, arguing that the replication rate is not significantly influenced by the presence or absence of potential flaws in these studies. Although continuing to disagree on the seriousness of the "flaws," the reviewers have agreed that "the present data base does not support any firm conclusion about the rela- tionship between 'flaws' and study outcome (Hyman & Honorton 1986). (Flaw analyses have yet to be reported on the REG and differential effect data bases.) The Hyman-Honorton ganzfeld debate is continuing in the Journal of Parapsychology. Whatever its final outcome, the discussion will lead to a more accurate interpretation of the data and better research in the future. In the final analysis, the case for psi cannot be won or lost by arguments over past experiments, but only by systematic and sustained new research that will survive the test of time. Honorton has recently reported con- tinued success using an automated testing protocol that would appear to answer Hyman's methodological objec- tions to the earlier ganzfeld research (Berger &- Honorton 1985; Honorton & Schechter 1986). 4.2.4. "Disbelievers" as replicators. Several critics of psi research (Alcock 1981; Kurtz 1981; Moss & Butler 1978) have argued that the replication work must be done by investigators who are unsympathetic to psi, a category that would exclude most (but not all) parapsychologists. Moss and Butler, for example, argue that "replication by a qualified nonsympathetic observer is the only guard against results which may have been 'contaminated by a conscious or unconscious bias" (p. 1068). We are now aware of its being common practice in other sciences to disqualify positive results from experi- ments conducted by researchers who are favorably dis- posed to the hypothesis they are testing. The personal beliefs of researchers are rarely reported and may often be difficult to determine reliably. We suspect, however, that if such a standard could be applied retrospectively to published research in psychology, for example, there would not be much left. The fact that parapsychologists are singled out for this treatment is symptomatic of the often ad hominem nature of the psi controversy. We have : CIA-RDP96-00789ROO2200110001-5 119:WAVIORAL AND BRAIN SCIENCES (1987) 10:4 547 nao oc raimer: Parapsychology review Approved For Release 2000/98/08 I -yet'to hear a critic suggest that negative results rorn "disbelievers" in psi be rejected on this basis. Although it is reasonable to assume that experimenters who obtained strong positive results in the first few psi experiments they conducted were converted to a "belief' in psi by these results (if they were not "believers" already), we have far too few data to draw any conclusions about the distribution of attitudes of investigators at the time they undertook their first psi experiments. Thus we really do not know how many "disbelievers" have ob- tained positive psi results. Finally, one cannot assume that confirmatory evi- dence, even from hardened "disbelievers , will neces sarfly be acknowledged as such. BBS readers might find it instructive in this connection to study what happened when certain members of the Committee for Scientific Investigation of Claims of the Paranormal quite unexpect- edly confirmed Michel Gauquelin's astrological "Mars Effect." (See Zetetic Scholar 1982a; 1982b; 1988; and references contained therein.) On the other hand, the fact that the outcomes of psi emeriments seem to b tive at least to AAc=e, to tbg-identity of the experim -nter or principal investigator is a legitimate cause for conce-r-n-.-nis-experimenter effe--cT in -Tpara~sych6logy has' 16-ng been recognized and extensively discussed within the field (e.g., Kennedy, J. E. & Taddonio 1976; White 1976a; 1976b); even some strong proponents of psi have had trouble obtaining positive results in their experiments. The jury is still out as to why this state of affairs exists. Until more is known, it is unwarranted and unfair to jump to the conclusion that the experimenter effect is due to fraud, negligence, or incompetence on the part of the successful experiment- ers, especially in the absence of supporting empirical evidence. The number of trained scientists who have obtained positive results in psi experiments is by no means inconsiderable, and many of these scientists have published in orthodox areas. More important, other plau- sible explanations of the experimenter effect can be proposed. For example, it is not implausible from a psychological point of view that an experimenter who does not expect positive results could convey this attitude to his subjects by nonverbal cues, thereby adversely affecting their confidence or motivation and thus their performance on the psi task. There is evidence from psychology for just such a process (Rosenthal & Rubin 1978). In addition, several studies within parapsychology that compared experimenters who had different attitudes or expectations about psi, or who behaved differently toward their subjects, have provided more direct support for this hypothesis (e.g., Honorton et al. 1975; Parker 1975; Taddonio 1976). The correct explanation(s) of the experimenter effect can come only from more research. This will come sooner if more scientists outside the parapsychological commu- nity - "believers," "disbelievers," and neutrals - can be persuaded to undertake psi experiments of their own, and to publish their results irrespective of outcome. Despite our remarks earlier in this section, we think that the involvement of a wider range of investigators in psi research is important and we wish to encourage such involvement. Indeed, that was one of our objectives in writing this BBS target article. We and other para- psychologi s would be asedio fonsult with any, Quali- ApproveYwor e ease 2000 OB/08 W -RDP96-R0789§0R22t001 e scientist w o wou d ni'doe luch an 4!p I e 0 R21 experiment. 5. Patterns, order,-and sense in parapsychology Has parapsychology gone any further than merely sug- gesting that anomalies exist? We think it has. Although some work in the field is still concerned with demonstrat- ing the integrity of the anomalies, emphasis in recent years has shifted strongly to so-called process-oriented research designed to uncover lawful regularities between psi and other psychological or physical variables. For example, there have been successful attempts to relate psi to subjects' beliefs and attitudes (Schmeidler & Mc- Connell 1958), personality and motivation (Eysenck 1967; Honorton & Schechter 1986), and to cognitive variables such as memory (Rao et al. 1977), visual imagery (Kelly et al. 1975), and stereotypy of responses to ESP target sequences (Stanford 1975). We would like to focus here, however, on one hypothesis that appears to bring together a large and diverse body of experimental results: the idea that psi may be facilitated by procedures that result in the reduction of meaningful sensory and pro- prioceptive input to the organism, and the concomitant redirection of attention to internally generated imagery. This hypothesis is known in parapsychology as the noise reduction model. Whatever its "real" mechanism, ESP may usefully be thought of as behaving like a weak signal that must compete for the information-processing resources of the organism. It follows that the reduction of ongoing sen- sorimotor activity may facilitate ESP detection by the organism. As illustrated in a book by the psychologist Harvey Irwin (1979), the noise reduction model fits in well with concepts that are widely accepted in cognitive psychology and information-processing theory. The model is particularly relevant to the notion of limits in the information-processing capacity of the organism (Kahne- man 1973); namely, the more internal and external .. noise" the system must process, the less is available to process possible psi information. It is interesting that most of the traditional techniques of "psychic" development seem to involve some form of reduced vigilance or "noise reduction. " For example, the practice of yoga, which is believed among other things to help develop ESP ability, appears to involve procedures that control habitual sensory, autonomic, and cognitive processes (Rao et al. 1978). The first five of the eight stages in Patanjah's yoga, for example, are preparatory and are aimed at achieving voluntary control of internal processes. The ability of yogins to exercise unusual con- trol over heartbeat and EEG activity, to cause sweat on certain parts of the body, and become physiologically nonresponsive to external stimuli has been satisfactorily documented (Anand et al. 1961; Wallace 1970; Wallace et al. 1971). The final three stages of yoga are dharana (concentration), dhyana (meditation), and samadhi (a state of stillness ofthe mind). If the introspective accounts of the yogins are any guide, the dharana state seems to involve intense focusing of attention on a single object, whereas meditation (dhyana) enables the practitioner to hold that focus over an extended period of time, which is believed to result in a stand-still state of mind (samadhi). CIA-RDP96-00789ROO2200110001-5 548 BEHAVIORAL AND BRAIN SCIENCES (1987) 10:4 Approved his state is also described as an expansion of con !iousness that goes beyond the object of perceptual :tention (Dasgupta 1930). There is voluminous phe omenological information on this, along with a modicum f physiological data (see, e.g., Das & Gastaut 1955). Historically, many of those who have claimed suc essful psi receptivity have also claimed that they did ,ieir best when they were physically relaxed and when he mind was in a "blank" state. Rhea White (1964), who eviewed the early literature on this topic, concluded that ttempts "to still the body and mind" are common among he techniques used by successful psi subjects. Mary ;inclair, whom her husband, Upton Sinclair, found to be in excellent psi subject, recommended for a successful 3si outcome that "you first give yourself a 'suggestion' to he effect that you will relax your mind and your body, snaking the body insensitive and the mind a blank" 'Sinclair 1930, p. 180). White (1964) further elaborated this technique and classified it into four stages: (1) relaxa tion; (2) engaging the conscious mind by keeping it blank or focusing on a single mental image or feeling, perhaps following this by a "demand" that the psychic impression come; (3) waiting patiently for the impression to appear; and (4) assessing rationally if the impression is psychic. There is also a large body ofexperimental evidence that procedures enabling a subject to limit extraneous sensory and proprioceptive input are conducive to the manifesta- tion of psi. Much of this evidence has been comprehen- sively reviewed by Honorton (1977), so we will limit ourselves to a brief discussion of work in five areas - ganzfeld stimulation, hypnosis, relaxation, meditation, and dreams. 5. 1. Ganzfeld and ESP The research on ESP in the ganzfeld has already been discussed at some length. One additional point may be added that is particularly relevant to the present discus sion: Those studies that assessed the self-reported effects of the ganzfeld on sub ects' state of consciousness have j generally found that the largest mean deviation scores from chance on the ESP test occurred among those subjects who claimed the greatest psychological effect from the manipulation (Palmer 1978; Sargent 1980). 5.2. Hypnosis and ESP There is an extensive experimental If terature on ESP and hypnosis. Fabler and Cadoret (1958), for example, tested college students in two formal experiments using a clair- voyance type of card-guessing task. In half of the trials the subjects were "under hypnosis" as they attempted to guess ESP cards screened from their view, and in the other half they guessed the targets while in a waking state. The order of testing was counterbalanced. In both experiments the subjects did significantly better in the hypnotic condition than in the waking condition. In a careful review, Ephraim Schechter (1984) evalu- ated data from 25 experiments in which ESP performance was compared in hypnotic and control conditions. The results of 5 of these experiments are uninterpretable for a variety of reasons. Of the remaining 20 studies, 16 show higher scores for the hypnotic condition, with 7 of them For Release 2000/08/08: showing statistical significance. None ofthe four reversals are significant. 5.3. Relaxation and ESP The most extensive work in this area has been carried out by William Braud. In one of the best designed of these studies (Braud & Braud 1974), 20 volunteer subjects were assigned randomly to "relaxation" or "tension" condi- tions. Those in the relaxation condition went through a taped, progressive-relaxation procedure (an adaptation of Jacobson's) before taking an ESP test, which was to guess the picture being "transmitted" by an agent in another room. The subjects in the other group were given taped, tension-inducing instructions before they did the same ESP test. Each subject's level of physical tension was assessed through electromyographic recordings and self- ratings. Both measures revealed a significant decrease in tension among the subjects in the relaxation group and a significant increase among those in the tension group; as predicted, the ESP scores of the subjects in the relaxation group were significantly above chance and significantly higher than those of the subjects in the tension group. Although no formal meta-analyses have been con- ducted on this data base, our own informal survey un- covered 13 series from six researchers that have reported significant effects (two-tailed) favoring the facilitative ef- fect of relaxation, and only one significant reversal using the same criteria. 5.4. Meditation and ESP Studies investigating meditation and psi suggest a positive relationship between these two variables. Rao et al. (1978) reported three series of experiments with a total of 59 subjects who had various degrees of proficien- cy in yoga and meditation. The subjects were given two ESP tests both before and after they meditated for at least half an hour. In one test the subjects "blind matched" cards with ESP symbols against target cards concealed in opaque black envelopes, and in the other test they attempted to describe concealed pictures. The results of both tests yielded independently significant premeditation-to-postmeditation dfferences when the three series were pooled. The card-testing results were also significant for each of the three series separately. Again, no formal meta-analyses have been conducted on this data base. However, our own informal survey uncovered 12 series from six researchers that have re- ported significant effects (two-tailed) favoring the facili- tative effect of meditation, and only one significant rever- sal, using the same criteria. 5.5. ESP in dreams Finally, mention should be made of a successful series of experiments on ESP in dreams conducted at Maimonides Medical Center (Ullman et al. 1973). In a typical experi- ment, a sender attempted to transmit the content of a randomly selected art print to a subject sleeping in an isolated room. When physiological monitoring indicated that the subject was dreaming, an experimenter blind to the target awakened the subject and elicited a dream report. The following morning, a tape of the dream Approved For Release 2000/08/08 : CIA-RDP96-00789ROO2200110001-5 BEHAVIORAL AND BRAIN SCIENCES (1987) 10:4 549 A at a0zyululogy review Aproved For M1 a%e 2A9P/08/08 reports was pla d Fack to the su ge o a itsso , , I P,w ciational material and a "guess for the night." Subse- quently, outside judges and/or the subject attempted to match the randomly ordered targets and dream tran- scripts from a series of sessions on a blind basis. In an article that appeared recently in American Psy- chologist, Irvin Child (1985) reviewed 15 separate series from the Maimonides program. After eliminating data from analyses that may have been compromised by non- independence of the judgings, he concluded that the remaining data were collectively significant both for the independent judges and for the subjects as judges. Child's article also documents several instances of gross misrepresentation of the Maimonides experiments in commentaries by critics. In contrast to the other research considered in this section, there have been no independent replications of the Maimonides research that have provided significant results. Two major failures to replicate have been re- ported (Belvedere & Foulkes 1971; Foulkes et al. 1972), and one other is equivocal (Globus et al. 1968). 5.6. Some criticisms Considering the legendary elusiveness of psi, the rate of reported success in the psi studies involving sensory noise reduction, although far from perfect, is impressive, even more so because the results appear to make sense in the context of both traditional psychic training practices and theories from orthodox psychology. One could of course point out that studies such as the so-called remote- viewing experiments (Targ & Puthoff 1977), which do not involve any explicit procedures for reducing sensory noise, have also recorded success rates of about 50%, arguing that our rationale is unsupported by these stud- ies. However, such an argument does not take into account the fact that most of the successful remote view- ing experiments, unlike the experiments discussed above, used subjects that were preselected for psychic talent and thus less likely than ordinary volunteers to need a supportive cognitive state to perform successfully. Second, there is reason to believe that at least some of these subjects attempted to reduce noise on their own. Marilyn Schlitz, a highly successful remote viewing sub- ject, put herself in a "calm state throughout," even though she used no formal relaxation procedure (Schlitz & Gruber 1980). Dunne and Bisaha (1978) asked their remote viewing subjects to "relax and clear their minds" prior to the remote viewing test. Even if one were to concede that successful remote viewers are generally in an ordinary state of conscious- ness during the psi task, it does not follow that they might not have performed even better had they been in an altered state of the type we have been discussing. This observation, however, brings to light another criticism of the studies supporting the noise reduction model. Many of these studies, in particular most of the ganzfeld and relaxation experiments, failed to use control groups or other means of assessing whether the induction pro- cedure was actually responsible for the positive scoring. Among those studies that did use such controls, the designs still did always preclude other interpretations of the results (see, e.g., Stanford 1987). Especially in the experimen'R usingitIhirpsubjrxts designs relative sue- ppr ve or Release 2000/08/08 -9 -Wf 'MMAJAes be eglih It P?4§n WMiU0A attributable to expectancy effects or demand character. istics. More research will be needed before the status of the noise reduction model can be cbnclusively determined. A large body of empirical data from diverse sources is nevertheless consistent with this hypothesis. This fact is sufficient to support the more modest point we are trying to make: Psi data fall into patterns that make psycho- logical sense and encourage a systematic program of re- search. 6. Practical significance The remaining criticism that needs to be addressed con- cerns practical significance. Even if one concedes that the preceding criticisms have been addressed satisfactorily, it can be argued that the results of psi experiments are trivial and of no practical or clinical importance. It is certainly true that the effect sizes in most psi experiments are small. For example, the effects reported by Schmidt in his REG experiments rarely exceed chance expectation by more than a few percent. Such outcomes hardly seem to be practically useful. There are fallacies in this line of criticism, however. First, it fails to acknowledge the distinction between basic and applied research. Practical significance is in- deed important if the objective is to determine whether a process can be applied to solve "real-world" problems. Parapsychology, however, is devoted almost exclusively to basic research, where the objective is to address theoretical issues. Psi results seem to violate expectations derived from generally accepted physical theory, and this makes them of theoretical interest irrespective of their magnitude. For example, many of the most important experiments in physics deal with effects of very small magnitude. The above criticism is problematic even from the applied perspective, however, because techniques from information theory can be implemented to amplify a weak effect of the type commonly found in psi experiments. In one experiment, for example, Ryzl (1966) had the subject Stepanek guess whether the green or white sides of 30 cards placed inside opaque envelopes were uppermost. The cards were rerandomized and Stepanek guessed the order again. This process was repeated until Stepanek's distribution of guesses on each of 10 principal cards favored either green or white to a prespecified degree. Other criteria involving the other. 20 cards also had to be met. The result was a single "majority vote" on each of the 10 principal cards. In each of five experiments, Step- anek's majority votes duplicated the target order of the 10 principal cards perfectly (100%), although his success rate on individual guesses was only 62%. Other examples of this approach have also been documented (e. g., Car- penter 1975; Puthoff 1985). The reason that psi has not yet been applied on a broad scale has to do not with the size of the effects but with their unreliability, which (as discussed above) probably reflects our lack of understanding of the fitctors that affect performance on psi tasks. Uncovering these factors is a prime objective of modem parapsychological research. if k~apomalies do in fact turn out to represent some CIA- 96-00789R002200110001-5 550 BEHAVIORAL AND BRAIN SCIENCES (1987) 10:4 an- Approved For Release 2000/08/08 : CIA-RDFVJV1 - itdfff 661416661% ~ret6fbre unrecognized and far-reaching ability to ac- LiUe information and manipulate the environment, and this ability could be brought under conscious control, ie practical applications and potential benefits to man- ind seem almost limitless. It is easy to put para- sychologists on the defensive by citing the slow progress lat has been made to date in coming to grips with the nomalies. What such an approach overlooks is the im- ortance of solving the admittedly unsolved puzzle that he anomalies represent, It seems to us that too many ommentators on both sides of the psi controversy place ,Xcessive faith in what amounts to little more than spec- ilations about the true nature of the anomalies. Only by -ontinued research, preferably supported in a mean- ngful way by the scientific community at large, will the 'peculations turn into knowledge. 7. Conclusion We find that the frequency of replications, especially with regard to the noise reduction hypothesis, indicates that we are indeed on the trail of something interestin Atthe 9. same time, we cannot totally rule out the possibility that we may yet discover a hidden artifact or set of artifacts that would provide a satisfactory conventional explana- tion of the results (and which might, in their own way, likewise prove interesting). Such an open approach, which is widely shared within the parapsychological com- munity (Parapsychological Association 1986), is dictated bv the anomalous nature of psi and the fact that there is shill no verified theory of the mechanism(s) involved in psi interactions. Scientists working in this area must accord- ingly approach all hypotheses with an attitude of skep- ticism and must show a readiness to look at various alternatives (Palmer 1986a). Critics with a great deal of a priori skepticism about psi have reasonable grounds for not accepting omegic hypotheses - that is, that e anomalies represent a new principle of nature. At the same time, they have little justification for choosing to close their minds to the alternative possibility - namely, that the anomalies might reveal a currently unrecognized human capacity of great potential importance. If they do close their minds, they make the same mistake as those "believers in the paranormal" who refuse to study evi- dence and arguments contrary to their beliefs. At the least, there is now an excellent prima facie case for the statistical repeatability of the anomalies under certain conditions. There appears to be a common thread running through these studies, diverse though they may be, in the techniques of eliciting and measuring psi. . This commonality appears, at least in a crude and preliminary way, to make some theoretical sense and is leading to work now in progress at various laboratories to refine and consolidate the methods and concepts. We have discussed here some experimental evidence for the reality of psi, as well as the objections of critics to such evidence. We have also considered the idea that sensory noise reduction may be favorable to psi, sketch- ing the experimental results that bear on this hypothesis. The following conclusions seem to emerge: (1) Schmidt's results and ma"ny other parapsychological findings would be taken seriously if they related to a conventional area in science, for standard methodological and statistical criticisms have been answered. (2) No single experiment, no matter how carefully designed and executed, can be expected to settle a controversial claim. The results of one good experiment do no more than make a claim. The significance of that claim is proportional to the degree that experiments supporting it are successfully replicated, and the degree of research and hypothesis-testing it generates. Also important is its potential for contributing to a theoretical understanding of the natural world and for practical application. (3) The issue of replication and the meaning of experi- mental results in psi research have been a primary con- cern of parapsychologists. The discussion of the studies bearing on psi and sensory noise reduction and the rationale behind them show (a) a moderately significant rate of replication (in a statistical sense) and (b) the possibility of finding conditions that favor or inhibit psi. Together, these studies make a strong prima facie case for a genuine scientific anomaly and provide a viable re- search program. (4) Further clarity and precision in the concepts and hypotheses are needed. Noise reduction, for example, needs to be defined more precisely. Some improvements in experimental design may have to be introduced to deal with the central issue of how psi operates. No mechanism or theory that would adequately explain psi has been validated. Those who accord an extremely low subjective probability to omegic hypotheses may therefore justifia- bly demand more and better evidence. But demanding such evidence is not the same as questioning the cred- ibility of past research. (5) The final settlement of the question of the status of psi will have to depend on further research. The scientific legitimacy of psi cannot be denied by personal innuendos and ad hominem arguments, just as it cannot be estab- lished by preaching. One can only hope that the climate of scientific opinion wfll be sufficiently tolerant to permit free and open inquiry .by those who have the necessary skills and interest. NOTE 1. The theoretical rationale of the study was that the subject could psychokinetically influence the selection of the random seed numbers retroactively. We will not elaborate this hypoth- esis further, as it is not directly relevant to the control features of the experiment. Approved For Release 2000/08/08 : C"1DPLMr-0M9R1M221D0"000f%%-