A Review of Dream ESP Studies Conducted Since the Maimonides Dream ESP Programme

We review the dream ESP studies conducted since the end of the Maimonides research programme. Combined effect size estimates for both sets of studies (Maimonides r = 0.33, 95% C.I. 0.24 to 0.43; post-Maimonides r = 0.14, 95% C.I. 0.06 to 0.22) suggest that judges could correctly identify target materials more often than would be expected by chance using dream mentation. Maimonides studies were significantly more successful (p< 0.05) than post-Maimonides studies, which may be due to procedural differences, including that post-Maimonides receivers tended to sleep at home and were generally not deliberately awakened from REM sleep. Methodological shortcomings of some studies are discussed. Nevertheless, home dream ESP research has been successful and continues to be a less expensive and less labour-intensive alternative to sleep-laboratory-based research. We hope that interest in dream ESP research will be re-awakened. This paper aims to review studies of alleged dream extrasensory perception (ESP) conducted since the end of the Maimonides research programme and to compare and contrast their respective methodology and success. As defined by Irwin (1999, p. 6), ‘An extrasensory experience is one in which it appears that the experient’s mind has acquired information directly, that is, seemingly without the mediation of the recognized human senses or the processes of logical inference’. ESP can be further classified: telepathy (information about the present obtained from another person); clairvoyance (information about present events or objects obtained from the environment); precognition (information about future events); retrocognition (information about past events). Many spontaneous case reports of alleged ESP occur while the experients are in some kind of altered state of consciousness (ASC) (see Alvarado, 1998). Case Journal of Consciousness Studies, 10, No. 6–7, 2003, pp. 85–109 Correspondence: Simon Sherwood, Division of Psychology, University College Northampton, Boughton Green Road, Northampton NN2 7AL, UK. Email: [email protected] collections suggest that a large proportion, as much as 65% (Rhine, 1981), of spontaneous cases of ESP have occurred during dreams. Clients undergoing psychological therapy have also experienced dream ESP (e.g., see Krippner, 1991; Van de Castle, 1977). Dreaming is an obvious ASC for researchers to focus on (Ullman & Krippner with Vaughan, 1973; 1989) because it is naturally occurring and contains features considered to be important facilitators of ESP (see Braud & Braud, 1975; Honorton, 1977). However, it has the disadvantage of being very time consuming and requires expensive EEG–EOG monitoring equipment and sleep laboratory facilities if participants are to be deliberately awakened from REM to report their dreams, and so researchers turned to using the Ganzfeld technique during the 1970s. This technique is less expensive, less labour intensive, and was believed to induce a state similar to the hypnagogic (HG) state, i.e., the state that is entered just as one is falling asleep, a state also considered to include psi-conducive features of the sleep state. Although the Ganzfeld has been the dominant paradigm for ESP research since then, and has provided some of the best evidence (e.g., Bem & Honorton, 1994; Bem et al., 2001 — but see Hyman, 1994; Milton & Wiseman, 1999), the extent to which it induces an ASC is not clear (Alvarado, 1998). In fact, in a recent study, Wackermann et al. (2000) concluded that ‘[C]ontrary to the common belief, the ganzfeld does not necessarily induce a true hypnagogic state, and will surely not do so in most ganzfeld settings’ (p. 302). The Maimonides Dream ESP Studies Psychiatrist Montague Ullman established a dream laboratory at the Maimonides Medical Center, Brooklyn, in 1962 (Krippner, 1991). Before the laboratory closed in 1978 (Krippner, 1991; 1993; Ullman et al., 1973; 1989), his research team had conducted thirteen formal dream ESP studies and three groups of pilot sessions (see Table 1). Of the thirteen formal studies, eleven were designed to investigate telepathy and two precognition. The pilot sessions were designed to investigate clairvoyance, telepathy and precognition, respectively (see Table 1) The Maimonides procedure was developed and improved over time and a number of different procedural variations were explored. Thus, the following is intended as only a general description of a trial designed to investigate telepathy. The receiver was attached to EEG–EOG monitoring equipment and slept in a sound-attenuated room in the laboratory. Once he or she was asleep, a target was randomly selected from among a set of targets (typically art prints), selected on the basis of emotional intensity, vividness, colour and simplicity. The target, in a sealed envelope, was given to the sender, who was then locked inside another sound-attenuated room in the building (or, in some studies, a different building). The experimenter monitored the receiver’s EEG–EOG throughout the night and, 86 S.J. SHERWOOD & C.A. ROE [1] A sensory habituation method that encourages the internal focusing of attention and minimizes sensory distractions while the participant is physically relaxed. [2] It is possible that clairvoyance and/or precognition could also have operated. once the receiver had entered REM sleep, signalled the sender (via a buzzer) to open the target envelope and begin sending the target. At the end of the REM period, the experimenter awakened the receiver via an intercom and asked him or her to describe any dream(s). Responses throughout the night and in the morning were tape-recorded and later transcribed. The sender heard the receiver’s dream report via a loudspeaker, which may have reinforced his or her subsequent sending strategy. The receiver then went back to sleep. This process was repeated for each REM period with the same target being sent each time. In the morning, the receiver reported any associations to the dream mentation and guessed what the target might be. Receivers typically viewed between eight and twelve pictures, one of which was the target, gave a confidence rating for each picture and also placed them in rank order according to the correspondence with their dream mentation, associations and/or guesses. Complete dream transcripts and target sets were also sent to independent judges who made similar judgements. The ratings/rankings from the two or three blind judges were combined. A trial was a ‘binary hit’ if the target picture had been ranked in the top half of the target set and a ‘binary miss’ if ranked in the bottom half. Performance was then evaluated to determine whether it was significantly higher or lower than mean chance expectations (MCE). During most of the telepathy studies (see Table 1, A–H) the receivers’ dreams were monitored and recorded throughout the night and the same target was sent during each REM period (Child, 1985). However, during two studies known as the ‘Sensory bombardment’ and ‘Grateful Dead’ studies (L, M), the sending periods did not occur regularly throughout the night and did not necessarily coincide with the receivers’ REM periods. In the study with A. Vaughan, I. Vaughan, Harris and Parise (study O), some trials involved sending a different target during each REM period. Studies using the same receiver across all trials often used the same sender, too (B, D, E, F), but not always. Successful sender and receiver pairings from the two screening studies (A, C) were used in later studies. Some studies used more than one sender (A, C, G, O), either across a series of trials with the same receiver or different receivers. There was not always a single sender for each receiver either; for some of the sensory bombardment (L) trials there was a single sender for two receivers; for the Grateful Dead trials a concert audience of about 2,000 people acted as senders. During precognition and clairvoyance trials there was no sender. The distance between the sender and receiver also varied across the studies (e.g., A & B vs. L & M). Some studies employed ‘multisensory’ targets rather than just static art prints: in the second study with Erwin (F), the sender was provided with objects related to the art prints and asked to act out aspects of the scenes; in the first study with Bessent (I), Bessent spent an hour the following morning looking at a picture and immersing himself in a multisensory environment that accompanied this; in the second Bessent study (J), the targets were slide sequences with accompanying DREAM ESP STUDIES SINCE THE MAIMONIDES PROGRAMME 87 [3] In the early studies (A–C — see Table 1) confidence ratings for the rankings were given on a five-point scale but from the Posin study (D) onwards a 100-point scale was used. soundtracks. Slide sequences with a soundtrack were also used in the sensory bombardment study (L). In the final Maimonides study (Honorton et al., 1975), films were used. Maimonides dream ESP success During his review, Child (1985) discovered that the only way the results could be analyzed across the Maimonides series was in terms of the number of binary hits and misses. For most studies, these data were available for both participants and independent blind judges but in some studies only blind judging was conducted (see Child, 1985, Table 1, studies F, I–K). Child used the data based upon judging of the whole dream transcripts (which included associations as well as the participants’ guesses). Child (1985) concluded that: The outcome is clear. Several segments of the data, considered separately, yield significant evidence that dreams (and associations to them) tended to resemble the picture chosen randomly as target more than they resembled other pictures in the pool. (p. 1223). A meta-analysis of 450 Maimonides ESP trials (based upon the blind judges’ data) found the overall success rate to be 63% (MCE = 50%) with odds of 75 million to 1 against achieving such a result by chance (Radin, 1997, pp.71–2). This meta-analysis also found that the binary hit rate for 20 of the 25 sets of data analysed was above the MCE. Statistical significance can provide an indication of the probability of obtaining such an outcome if the null hypothesis were true but it can not provide an indication of the magnitude of the effect. Whether or not a statistical test produces a significant outcome will depend upon the magnitude of the effect, the power of the test and the sample size. Conversion of a test statistic to a common effect size measure has the advantage over conventional significance testing in that it provides an indication of the magnitude of any effect and permits direct comparisons across studies with different sample sizes (Prentice & Miller, 1992). Thus, we have converted the test statistics for the judges’ ratings/rankings (either z or t values) into an effect size measure r (see Clark-Carter, 1997, pp. 550–1, 558) for the twelve formal studies and three pilot studies listed in Table 1 of Child’s (1985) review. A positive effect size indicates that performance was above chance expectations; a negative effect size indicates that performance was below chance. Cohen’s (1977) rule of thumb suggests that r = 0.1 would be considered a small effect, r = 0.3 a medium effect and r = 0.5 or above a large effect. For the fifteen sets of data, the effect size r ranges from –0.22 to 1.10 (see Table 1). Interestingly, the studies with the largest effect sizes mostly involved 88 S.J. SHERWOOD & C.A. ROE [4] The correlation coefficient r is one of the most commonly used effect size measures (Prentice & Miller, 1992). [5] The Honorton et al. (1975) study was not included in Child’s review and cannot be included here because complete statistical test results are not available. [6] As with other correlation coefficient estimates, r should fall in the range –1 to +1. However, where z scores are large, it is possible for r to exceed this range. gifted single participants who had been pre-selected (i.e., Erwin, Van de Castle, and Bessent) and two of the least successful Maimonides studies were the two screening studies (A, C) that identified successful senders and receivers for use in subsequent studies. The two precognitive studies and one pilot study (I–K) were very successful (effect size ranges from 0.47 to 0.73). The clairvoyance pilot sessions (N) were also successful but less so (r = 0.35). The most successful Maimonides dream ESP study (r = 1.10) was the sensory bombardment telepathy study (L); other studies that employed multisensory targets were also very successful (F, I, J) (r = 0.65 to 0.88). Criticisms of the Maimonides studies The main criticisms of the Maimonides studies are concerned with the lack of replication (see Hyman, 1986; Krippner, 1991; Parker, 1975, p. 90) and the number of statistical analyses, and by whom these were conducted (Child, 1985; DREAM ESP STUDIES SINCE THE MAIMONIDES PROGRAMME 89 Study Type of ESP Trials Test statistic Effect size r A Ullman et al. (1966) study 1, first screening Telepathy 12 z = 0.71 0.205 B Ullman, et al. (1966) study 2, first Erwin study Telepathy 7 z = 2.53 0.956 C Ullman (1969), second screening Telepathy 12 z = –0.25 –0.072 D Ullman (1969), Posin study Telepathy 8 z = 1.05 0.371 E Ullman et al. (1973), Grayeb study Telepathy 8 z = –0.63 –0.223 F Ullman & Krippner (1969), second Erwin study Telepathy 8 t = 4.93 0.881 G Krippner & Ullman (1970), Van de Castle study Telepathy 8 t = 2.81 0.728 H Pilot sessions Telepathy 67 z = 4.20 0.513 I Krippner, et al. (1971), first Bessent study Precognition 8 t = 2.81 0.728 J Krippner et al. (1972), second Bessent study Precognition 8 t = 2.27 0.651 K Pilot sessions Precognition 2 z = 0.67 0.474 L Krippner et al. (1971), Sensory bombardment study Telepathy 8 z = 3.11 1.100 M Krippner et al. (1973), Grateful Dead study Telepathy 12 z = 0.61 0.176 N Pilot sessions Clairvoyance 8 z = 0.98 0.346 O Honorton et al. (1972), Vaughan, Harris, & Parise study Telepathy 203 z = 0.63 0.044 Honorton et al. (1975) Telepathy — unknown unknown