The Mind Machine

For many years scientists have examined the possible existence of extra-sensory perception (ESP). One of the most common types of experiment, referred to as a ‘forced choice’ study, involves participants attempting to guess the identity of hidden targets that have been randomly selected from a set of alternatives known to participants prior to making their guess. Many researchers have argued that the results of these experiments provide strong support for the existence of psychic ability. However, others have criticised many of the experiments on both methodological and statistical grounds. The authors aimed to help resolve this debate by devising a novel way of carrying out a large scale forced choice ESP experiment. The Mind Machine consisted of a specially designed steel cabinet containing a multi-media computer and large touch screen monitor. The computer presented participants with a series of videoclips that led them through the experiment. During the experiment participants were asked to complete a forced choice ESP task that involved them guessing the outcome of four random electronic coin tosses. All of their data was stored by the computer during an eleven month tour of some of Britain’s largest shopping centres, museums and science festivals. 27,856 participants contributed 110,959 trials, and thus the final database had the statistical power to detect the possible existence of a very small ESP effect. However, the cumulated outcome of the trials was consistent with chance. The experiment also examined the possible relationship between participants’ ESP scores and their gender, belief in psychic ability and degree of predicted success. The results from all of these analyses were non-significant. Also, scoring on ‘clairvoyance’ trials (where the target was selected prior to the participant’s choice) was not significantly different from ‘precognitive’ trials (where the target was chosen after the participants had made their choice). Competing interpretations of these findings are discussed, along with suggestions for future research. Acknowledgments The authors would like to thank The Committee On The Public Understanding of Science, The Perrott Warrick Fund, Geomica and Barcrest for supporting the work described in this paper. We also wish to acknowledge the assistance provided by John Bain, Adrian Owen, Tina Gutbrod, Caroline Watt, Mike Hutchinson, Susan Blackmore and the staff at each of the venues who were kind enough to host the Mind Machine. Introduction Parapsychologists have carried out a large number of studies examining the possible existence of extra-sensory perception (ESP). One of the most principal types of experimental design uses the ‘forced choice’ procedure, in which participants are asked to guess the identity of hidden ‘targets’ (e.g., the colour of playing cards) that have been randomly selected from a set of alternatives known to participants prior to making their guess (e.g., they are told that the cards will be either red or black). Many of the early forced choice ESP experiments were conducted by Rhine and his colleagues at Duke University in the early part of the last century (see Pratt, Rhine, Smith, Stuart & Greenwood, 1940/1966). The majority of this work involved participants attempting to guess the order of shuffled packs of cards carrying the image of a star, circle, square, cross or wavy lines. These studies were often very labour intensive, and involved data collection and analysis being carried out by hand. Recent research has tended to use more automated procedures. For example, Schmidt (1969) developed an electronic device that randomly selected one of four lamps, prompted participants to indicate which lamp they thought the device had selected and provided feedback by lighting the target lamp after they had registered their choice. Similarly, Honorton (1987) developed ‘ESPerciser’ a computerbased system that presented participants with four on-screen boxes and asked them to guess which one had been randomly selected by the computer. Both systems automatically recorded information about both the selected targets and participant choices. These experiments have investigated a wide range of hypotheses (see Palmer, 1978, for a review). Some studies have examined the possible existence of telepathy by having another person, referred to as a ‘sender’, concentrate on targets prior to the participant’s guess. Other work has investigated clairvoyance by having participants attempt to guess the identity of targets that are not known to anyone else (e.g., the order of a deck of cards that have been shuffled and immediately sealed in an envelope). A third set of studies has examined participants’ precognitive abilities by, for example, having them predict the order of a deck of cards, shuffling the cards and then comparing the predicted order with the actual order. These studies have also examined how ESP scores are affected by different kinds of target material (e.g., symbols vs words), experimental procedures (e.g., providing feedback to participants about their scores vs not providing feedback) and individual differences (e.g., those that believe in ESP vs disbelievers). Many researchers have argued that the results of these studies support the existence of ESP. For example, Pratt et al. (1940/1966) reviewed the findings from more than 3.6 million guesses made in over 140 forced choice ESP studies conducted between 1882 and 1939. Many of the studies were independently significant and, as a group, provided strong evidence of above chance scoring. Likewise, the automated experiments conducted by both Schmidt (1969) and Honorton (1987) produced highly significant results. Also, Honorton and Ferrari (1989) presented a meta-analysis of nearly two million guesses from precognitive forced choice experiments conducted between 1935 and 1987. Although the cumulated effect size was small (0.02), the results were highly significant (p<10). More recently, Steinkamp, Milton & Morris (1998) carried out a meta-analysis of 22 forced choice ESP studies that had compared scoring between clairvoyant and precognitive trials. The cumulated outcome of both trial types was highly significant (precognition trials; effect size=0.01, p=9x10: clairvoyant trials; effect size=0.009, p=.002). The results of many experiments also contain evidence of significant internal effects, with, for example, participants who believe in ESP tending to outperform disbelievers (Lawrence, 1993), and trials employing feedback obtaining better results than those giving no feedback (Honorton and Ferrari, 1989). However, many forced choice ESP studies have been criticised on both methodological and statistical grounds. For example, Hansel (1980) and Gardner (1989) have claimed that some of the early card guessing experiments employed procedures that would have allowed for participant cheating and ‘sensory cueing’ (i.e., participants inadvertently detecting and utilising subtle signals about the identity of targets). Others have suggested that the highly significant results from some of the automated studies are invalid, as the experiments used nonrandom methods to select and order targets (see, e.g., Hyman, 1981; Kennedy, 1980). Critics have also pointed to possible problems with the way in which data has been collected and analysed. For example, both Leuba (1938) and Greenwood (1938) discussed the potential dangers of ‘optional stopping’, wherein researchers are able to conclude an experiment when the study outcome conforms to a desired result. Other studies have suffered from the ‘stacking effect’ a statistical artefact that can occur when guesses from many participants are all matched against the same target material (see, e.g., Pratt, 1954). The poor levels of methodological and statistical safeguards present in some past studies have been highlighted in two recent meta-analyses. Honorton and Ferrari (1989) analysed the quality of each study in their database by assigning a point for each of 8 major methodological criteria. The studies received an average rating of just 3.3. Likewise, in their meta-analysis, Steinkamp et al (1998) assigned clairvoyant studies a methodological quality rating of between 1 and 19, and reported that the studies obtained a mean rating of just 10.6. In reply, some parapsychologists have questioned the validity of these criticisms by, for example, arguing that many of the alleged flaws are unlikely to account for reported effects (see, e.g., Palmer, 1986a for a summary), and noted the non-significant correlations between the presence/absence of methodological safeguards and study outcome (see, e.g., Honorton & Ferrari, 1989; Steinkamp et al, 1998). The authors aimed to help resolve this debate by devising a novel procedure for carrying out a large scale forced choice ESP experiment. The Mind Machine consisted of a specially designed steel cabinet containing a multi-media computer and large touch screen monitor. The computer presented participants with a series of videoclips that led them through the experiment. During the experiment, participants were asked to complete a forced choice ESP task that involved guessing the outcome of four random electronic coin tosses. All of their data was stored by the computer during an eleven month tour of some of Britain’s largest shopping centres, museums and science festivals. This methodology was developed for several reasons. First, the study had the potential to collect a huge amount of data from thousands of participants, and thus possess the statistical power to reliably detect the small effect sizes reported in many previous forced choice ESP studies. Second, it is widely acknowledged that carrying out large scale forced choice ESP experiments is usually problematic, as they tend to be time consuming and tedious for both experimenters and participants alike (Radin, 1997: Broughton, 1991). The Mind Machine overcame these problems by creating a totally automated experiment and by having each participant only contribute a very small number of trials. Third, as noted above, many previous forced choice ESP studies have been criticised on various methodological grounds, including sensory shielding, opportunities for participant cheating and poor randomization. The Mind Machine was designed to minimise these potential problems. For example, the computer running the experiment was secured inside a locked cabinet that could not be accessed by participants. Also, possible randomization problems were minimised by having the target selection carried out by a pseudo-random number generator that had been fully tested prior to use. Fourth, again, as noted above, critics have correctly noted that some previous forced choice ESP studies have suffered from potential statistical problems, including optional stopping and stacking effects. The Mind Machine was designed to overcome these artifacts by specifying the size of the final database in advance of the experiment, and generating a new target sequence for each participant. Fifth, the Mind Machine methodology could incorporate many of the factors that have positively correlated with study outcome in meta-analyses of previous forced choice studies. Incorporating such ‘ESP conducive’ procedures was important, given that a previous metaanalysis of forced choice ESP studies conducted via the mass media under non-conducive conditions, had resulted in a cumulative outcome consistent with chance (Milton, 1994). Honorton and Ferrari’s (1989) meta-analysis of precognition forced choice studies noted that several factors were significantly associated with increased ESP scoring. Studies providing immediate, trial by trial, feedback to participants obtained significantly higher effect sizes than those giving delayed or no feedback. Also, experiments testing participants individually had significantly higher effect sizes than those employing group testing. Many of these patterns were also found in the meta-analysis carried out by Steinkamp et al. (1998). To maximise the potential of obtaining evidence for ESP, the Mind Machine tested participants individually and provided them with immediate, trial by trial, feedback. On the basis of the results from previous forced choice ESP studies, it was predicted that participants’ overall ESP scores would differ significantly from mean chance expectation. The Mind Machine also attempted to replicate one of the most reliable internal effects in the forced choice ESP literature. A meta-analysis carried out by Lawrence (1993) revealed that participants who believed in psychic ability tended to obtain significantly higher forced choice ESP scores than disbelievers. To examine this hypothesis, participants in the present experiment were asked to indicate whether they believed in the existence of psychic ability prior to completing the ESP task. On the basis of previous work it was predicted that the ESP scores of believers would be significantly higher than disbelievers’ scores. The study also investigated four other internal effects that have received mixed support in past research. As noted above, several studies have examined potential differences between ESP scoring on clairvoyant trials (i.e., where the target is selected prior to the participant’s choice) and precognition trials (i.e., where the target is chosen after the participants have made their choice). These studies have obtained mixed results. Although a few studies have reported precognition trials outscoring clairvoyance trials (see, e.g., Freeman, 1962; Honorton, 1987), the meta-analytic review conducted by Steinkamp et al. (1998) showed no significant difference between the two types of trials. The Mind Machine followed up on this work by comparing the scoring of clairvoyant trials (where the outcome of half of trials was determined prior to participants indicating their guess),with precognition trials (where the outcome of the other half were determined after they had indicated their decision). Previous forced choice experiments have also examined the potential relationship between ESP scoring and gender. Again, these studies have obtained mixed results, with some reporting females outperforming males, and others showing no significant differences (see, Palmer, 1978 for a review). The Mind Machine experiment examined the potential relationship between gender and ESP scoring. A small number of past forced choice studies have investigated whether participants’ predictions about their performance on a forthcoming ESP task is related to their actual performance. Some of these experiments have shown a significant relationship between predicted and actual performance (see, e.g., Smith, Wiseman, Machin, Harris & Joiner, 1997: Musso, 1965; Schmeidler, 1971) but others have failed to obtain this effect (see, Palmer, 1978 for a review). Participants in the Mind Machine were asked to predict how many coin tosses they believed they would guess correctly, and the relationship between their predicted and actual success examined. The experiment also explored whether participants taking part in the study for the first time would obtain significantly different ESP scores to those repeating the experiment. The small number of forced choice studies that have examined participant’s ESP performance over several sessions have tended to report a decline effect (see, e.g., Honorton & Carbone, 1971, Humphrey, 1945). At the start of the Mind Machine experiment participants were asked to indicate whether this was the first time they had taken part in the experiment, and thus it was possible to examine the ESP scoring of ‘novices’ to participants who had taken part in the study before.