Explorations with the Perceptual ESP Test.
Immediately prior to the presentation of the letter slide in this earlier work, a subliminal picture was flashed to the subject. The picture was either the standard Defense Mechanism Test (DMT) slide, in which a threatening male figure is peering down at a boy seated at a desk (Kragh & Smith, 1970), or a control slide in which the countenance of the male figure in the background is pleasant and smiling. Palmer and Johnson expected more negative ESP scores when the face was threatening than when it was smiling. The hypothesis was not confirmed, but two post hoc findings suggested that there was some ESP in the data. First, there was a confirmation of Stanford's (1975) response-bias hypothesis. When a particular letter was called six times or less out of the 40 trials, the number of hits on these undercalled trials was significantly above chance (p [is less than] .01) and significantly greater than on the remaining trials (p [is less than] .01).
Because one purpose of the perceptual ESP test is to minimize linear thought processes, it was hoped that there would be less evidence of response biases with this procedure than with a traditional forced-choice ESP test. It seemed that such biases were indeed reduced, particularly the undercalling of XX doublets, but the biases were still present to some degree. This was good news in one sense because it suggested that the perceptual responses to the letter slide were not merely random: there was evidence of some kind of information in the responses, whether or not any of that information was psychic.
The second finding was overall tight variance (p [is less than] .005), which has been associated with a negative mood in the literature (Rogers, 1966, 1967). The puzzling aspect of this finding was that the tight variance was present with the smiling subliminal picture as well as the threatening one. Two possible explanations were offered. First, the smiling picture may not have been as benign as was originally thought; for example, in retrospect, it could be construed as lecherous, especially by females. Second, a possibly negative context of the experiment as a whole could have caused the fight variance. For instance, to fulfill the requirements of a visual field manipulation, subjects had to wear uncomfortable goggles and place their chins on a chin rest during the ESP task. A third possibility was that the presentation of subliminal stimuli might itself have had a negative effect.
A partial replication of this experiment that excluded the visual field manipulation and the DMT control slide was subsequently conducted at the Institute for Parapsychology (Palmer, 1992). The targets were changed from the letters a to d to typewriter carets (or [caret]s) pointed up, down, left, or right. The test procedure henceforth was dubbed the caret test. The carets were chosen over the letters because they were thought to be more uniform physically and thus might further reduce response biases. They were also expected to have fewer semantic associations than the letters, especially for students, for whom the letters correspond to academic grades.
The significant psi-hitting on counterbias responses was not confirmed, but there was an encouraging trend in the direction of tight variance (p = .072, one-tailed). Post hoc analyses revealed that the tight variance was significantly more pronounced among subjects who did not characterize the DMT slide as hostile following full exposure (p = .012) and among those who reported relatively great difficulty in taking the caret test (p = .008). Subjects who reported one or more spontaneous impressions of the target before the caret slide flashed scored significantly above chance (p = .011) and significantly higher than the other subjects (p = .017). Another replication is currently under way using a new DMT control slide.
Finally, an exploratory experiment was undertaken in which the caret slide was easier for the subjects to see (16 white carets on a black background). The message YOU ARE SAFE HERE was flashed subliminally to the subjects four times before the one 40-trial run, and subjects listened to background music of Gregorian chants during the test. The caret slide was flashed for either 100 ms or 300 ms (Palmer, in press).
The results revealed overall psi-missing (p = .028), which was attributable to the subjects who expressed the least posttest confidence that they had demonstrated ESP in the experiment (p = .009). A curious interaction was found, whereby subjects' reported difficulty in taking the caret test was correlated positively with ESP scores in the 300-ms condition (p = .006), but negatively in the 100-ms condition. The correlations differed significantly from each other (p = .003).
In the present experiment, which actually was conducted prior to the last two experiments already described (Palmer, in press), it was decided to pursue the rationale of the DMT manipulation, but to do so with the relevant stimuli directed to the agent rather than the percipient. For half of the subjects, there was a faint line-drawing of a monster in back of the target caret that the agent "sent" on each trial. Although the monster was not strictly subliminal, it may have had the same effect because the agent could not attend to it: the slide was flashed for only 100 ms and the agent had to focus on the caret, which she needed to send and record.
Subliminal stimuli were also presented to the percipient before each run. A recent meta-analytic literature review by Hardaway (1990) has supported claims by Lloyd Silverman that repeated presentation of the subliminal message MOMMY AND I ARE ONE has various kinds of beneficial effects on psychiatric patients as well as on normal subjects. Silverman interpreted these data from a psychoanalytic perspective, but this perspective has not been proven to provide the correct explanation for the effect. Hardaway reported weaker but still statistically significant efficacy for other oneness stimuli, such as MY PROF AND I ARE ONE positively affecting examination grades of college students (Parker, 1982).
Because telepathy is sometimes construed as a kind of mental merging of agent and percipient, it was decided to try as the oneness message ASHLEY AND I ARE ONE, Ashley being the first name of the agent in the experiment. The agent also conducted the first (preliminary) experimental session, so that there was ample time for agent and percipient to get to know each other.
There are two control conditions that ideally should be used to evaluate the effect of a subliminal stimulus. The first is the supraliminal presentation of the stimulus. The second is the subliminal presentation of a structurally similar but semantically neutral stimulus. Silverman's standard subliminal control stimulus was PEOPLE ARE WALKING, but some critics (e.g., Balay & Shevrin, 1988) have suggested that it may not be truly neutral. Therefore, I chose a nonsense message, which consisted of replacing each letter of ASHLEY AND I ARE ONE with the letter X.
Only the supraliminal control condition was used in the first experimental series. It was planned that the subliminal control condition would be substituted for the supraliminal control condition in the second series if the results were promising. Thus, in the first series, half of the subjects were exposed to ASHLEY AND I ARE ONE subliminally and half, supraliminally.
The rationale for the perceptual ESP test implies that a dissociated state should facilitate psi in such a test. The idea is that a subject's conscious attention is focused on something else, whereas more involuntary or automatic eye movements govern where on the caret slide the subject fixates. An attempt was made to promote this kind of process by having subjects listen to baroque music during the ESP task and by encouraging them to focus their attention as much as possible on the music rather than on the psi task. An attempt was also made to promote this state by having long intertrial intervals.
This line of reasoning led to the expectation that subjects who frequently have dissociative experiences would be better able to get into a dissociative state during the ESP task than other subjects, and thus they would show more psi on the ESP task. Perhaps the best known paper-and-pencil measure of dissociative tendencies is a scale by Bernstein and Putnam (1986). However, an alternative scale developed by Riley (1988) was chosen because the items seemed less pathological in content than those on Bernstein and Putnam's scale. Such items seemed more appropriate because the dissociative behavior of interest in this experiment was not of a degree that would be considered pathological.
Subjects were also given a widely used measure of anxiety, the Spielberger Trait Anxiety Scale (Spielberger, 1983). An anxiety scale was included because of the monster manipulation. If subjects in the monster condition picked up psychically on the monster being flashed to the agent, their unconscious response to this stimulus, and hence their ESP score, might be mediated by their disposition toward anxiety. Because the anxiety was expected to be unconscious, it was decided that a trait measure would be more likely to accurately measure the anxiety response than a state measure, even though the latter would be more direct. From the literature, one can make a case for either a positive or negative relationship between anxiety and ESP scores in this situation, although at the outset my expectation was for a negative one. These considerations will be elaborated upon in the Discussion section.
Finally, there was an 11-item rating scale that the subjects filled out partly just before and partly just after the ESP task. The scale was designed to measure the subjects' general mood and expectations, as well as their reactions to the caret test and the broader test procedure. The item that is most directly relevant to the rationale of the experiment was how dissociative an approach the subjects took to viewing the caret slide. This item will be discussed more fully in the Results section.
The preceding theoretical rationale provided general ideas about the sorts of relationships to be expected in the data, but it was not refined enough to justify formal hypotheses. Thus, it was decided to comb the results of the first experimental series looking for theoretically relevant significant effects. If enough of them were found to make continuation of the project seem worth while, these effects would be turned into formal hypotheses for the next series. It was also planned to check whether the fight variance and the response-bias effects found by Palmer and Johnson (1991) showed up again, but the procedures in these two experiments were sufficiently different that the present experiment was not considered to be an attempted replication of the earlier one.
The main experimenter was J. P., whose principal role was to test the subjects in the main experimental session. The other experimenter was Ashley Ellington (A. E.), a junior psychology major at Duke University. Her main roles were to conduct the preliminary test session with the subjects and to serve as agent in the main session.
The number of subjects was set in advance at 40. They were recruited through ads in the Duke University student newspaper, talks on parapsychology given by J. P. to student dormitory groups, and personal contact by A. E. At this stage, subjects were told only that the experiment involved ESP and subliminal perception. Those who inquired further were told that the idea was to see if subliminal messages could influence scoring on an ESP task in which responses were based on eye fixations. Fifteen of the subjects were friends or relatives of A. E. There were 28 females and 12 males.
There were three test rooms, all located on the second floor of the Institute for Parapsychology. The subject was seated in a reclining chair in Room 1. J. P. was located in Room 2, which is adjacent to Room 1 and separated from it by a door and a 60-mm x 45-mm double-glass window. Room 2 housed the projector and computer. The agent (A. E.) was located in Room 3, which was separated from Rooms 1 and 2 by a hallway. The distance between subject and agent was approximately 7 m. The doors of all rooms were closed during the main test sessions. There was two-way intercom communication between Rooms 1 and 2 but not between either of these rooms and Room 3.
Equipment and Specific Layout
Stimuli were presented to the subjects by means of a two-channel projection tachistoscope (Gerbands Model G1170). Each channel consisted of a standard Kodak Ektagraphic III B slide projector fitted with a tachistoscopic shutter. One of these channels was set up in Room 2 and the other in Room 3. Both channels were controlled by an IBM-PC-type microcomputer and custom-made interface hardware located in Room 2.
The projector was in Room 2. Using a 250-W halogen lamp and a low brightness setting, it presented visual stimuli to the subject in Room 1. The stimuli, reproduced on slides, first passed through two pieces of polarizing filter, the axes of which subtended at an angle of 72 [degrees]. The filters were pasted directly on the shutter housing. Images next passed through the double window into Room 1 and then through a 12-mm-thick ground-glass screen mounted on a stand inside the room. The double window was covered in Room 1 by a wooden screen into which a 9-cm x 12-cm rectangle had been cut. The screen was covered by a piece of black cardboard into which had been cut a 12-cm x 14-cm rectangle which defined the area where the stimuli were exposed. The subject was seated 2 m in front of the screen, except as noted hereafter.
The projector in Room 3 presented stimuli to the agent. The stimuli were flashed onto a sheet of white paper mounted on the wall approximately 80 cm from the agent at eye-level. The agent was seated at a desk immediately to the right of the projector. A cardboard box was placed over the carousel to shield its movements.
A cassette tape deck in Room 2 funneled auditory stimuli into Room 1 through padded headphones worn by the subject. These stimuli were monitored by J. P. in Room 2 through a separate set of headphones attached to a second audio cassette tape deck connected to the first one.
Slides and Targets
All physical stimuli were slides made from photographing sheets of white paper onto which had been rubbed black symbols (Geotype GS-106) or onto which had been typed capital letters. The slides were darkened by photographing them at a high F-stop (F-11), using black-and-white film.
The test stimulus for the subject consisted of a display of 28 carets arranged in staggered rows, so that there were seven of the carets pointing in each of four directions (up, down, left, and right). The different directions were distributed over the slide in a quasi-Latin-square format that resulted in each of the directions being represented in all areas of the slide.
The target stimuli projected to the agent consisted of two sets of four slides. Each slide consisted of a single caret pointed in one of the four directions, with each direction occurring once in each set. In one set, each caret transparency was superimposed on a photographic transparency of a line drawing of a monster. Thus, the monster provided a background for the carets in this target set.
All the targets for the experiment were selected by a pseudorandom algorithm implemented by Turbo-Basic software. The sequence was seeded by the number of microseconds between two keyboard button presses. The total number of targets generated was 2,112, which exceeds the number of designated trials by 192. The extra trials were added so that if a particular run had to be aborted, a fresh target sequence could be substituted.
The total sequence was evaluated by chi-square tests for singlet bias, singlet bias summed over runs, doublet bias, and bias of XX doublets. These analyses were performed by the computer without exposing the individual targets. The system was set up in such a way that if any of the chi-square values approached or reached significance, a new set of targets could readily be generated. The chi-square values for the sequence chosen were all well within chance limits.
The four-target slides were placed in slots 1, 3, 78, and 80 of the carousel atop the projector in Room 3. The location of each slide corresponded to the code in the computer software. For each trial, the computer instigated a sequence of three forward and/or backward turns of the carousel, after which the slide corresponding to the chosen target direction was dropped into the projector. The timing of these movements and the return to the zero position after the trial was over were carefully calibrated so that the subject could not infer the target by noticing possible differences in the duration of the trials: Trial durations were precisely equated.
The other stimulus was a message containing the words ASHLEY AND I ARE ONE that was flashed subliminally (2 ms) or supraliminally (2 s) to the subject prior to each run. The 2-ms rate was chosen because pilot testing indicated that the stimulus could not be identified at this exposure speed even if the person leaned forward in the chair.
The Caret Test
The procedure for the caret test was as follows. The subject initiated the first trial by pressing a green button in the middle of the response box, which rested on the subject's lap. The subject began each trial with eyes closed. After 17 s, the subject experienced a tactile/auditory stimulus provided by a padded metal vibrational device (pulsator) placed on the left armrest of the reclining chair and on top of which the subject's open left-hand palm rested downward during the test. This stimulus was a cue for the subject to open his or her eyes and look at the screen, but without focusing on any particular part of it. Three seconds later, the caret slide flashed (supraliminally) for 100 ms. The subject then chose an ESP response that corresponded to whichever particular caret was most salient in the visual field when the slide flashed. For example, if the most salient caret was pointing upward (irrespective of its location on the slide), the subject would press the "up" button on the response box. If the subject missed the flash or for some other reason was unable to make a choice, a pass option was available: By pressing a red button on the response box, the subject could cause the trial to recycle with the target remaining the same. After making the ESP or pass response, the subject closed his or her eyes and the next trial began automatically. The end of the run was signified by a sequence of two pulses and the lighting of the red button by itself.
Design and Test Parameters
The ESP test consisted of two 24-trial runs. There were two manipulated independent variables. The first independent variable was whether prior to each run the subject observed the stimulus ASHLEY AND I ARE ONE subliminally or supraliminally. The second independent variable was whether or not the target stimuli flashed to the agent had the drawing of the monster in the background. These variables were incorporated in a 2 x 2 factorial design with 10 subjects assigned to each cell. Assignments were made by computer at the same time the targets were generated. For this the computer used a random permutation method based on the same algorithm used for target generation, but with a different seed. J. P. was generally blind to each subject's assignment to both independent variables until the session was completed, and A. E. was generally blind to the subliminality condition.(1)
Session 1. The subject was greeted by A. E. and briefly introduced to J. P., if he was available. The subject was then accompanied by A. E. to Room 1 and was seated in the reclining chair. A. E. explained the experiment to the subject, who then signed a consent form. After the light in Room 1 was dimmed, A. E. exposed the caret slide to the subject supraliminally and demonstrated the pulsator. This was followed by several practice trials on the caret test with A. E. present in the room to be sure the subject understood the procedure. Then the subject completed 40 additional practice trials with A. E. not present in the room. The stated and actual purpose of these trials was to have the subject "overlearn" the response procedure so that it would be as effortless and automatic as possible. No mention was made of targets for these trials, and in fact there were none. The interval before the pulse was reduced from 17 s to 7 s, so that the task lasted about five minutes.
Next, A. E. flashed the words I HAVE ESP to the subject 10 times at 3 ms. After each exposure, the subject reported what he or she saw. It was assumed that if a subject could not identify this message, which was simpler than the test message and was presented at a rate 1 ms slower, it is unlikely that he or she would be able to identify the subliminal message in the actual ESP session. (This assumption, of course, was tested directly in Session 2.)
Only one subject was able to (partly) identify the slide at this shutter speedy Afterwards, the slide was exposed supraliminally to be sure subjects could identify it at full exposure. Exposures began at 1 s and increased by .5 s until recognition occurred. Seven subjects required exposures of longer than 2 s to identify the slide. In these cases, the chair was moved forward slightly until the slide could be identified at 2 s. This distance was retained for the main session.
Finally, each subject completed the two personality questionnaires: the 26-item Questionnaire of Experiences of Dissociation (QED) and the trait portion of Spielberger's State-Trait Anxiety Inventory (STAI), Form Y (Spielberger, 1983). A time was then scheduled for the subject to return for the main session.
Session 2. The subject was greeted by J. P. and A. E., offered a cup of coffee or a Coke, and accompanied to Room 1. After a few minutes of friendly conversation, J. P. explained the procedure to the subject. It was mentioned that prior to each run either a subliminal or a supraliminal message would be flashed on the screen four times. If it was subliminal, it would be a rapid flick; and if supraliminal, it would last two seconds. The subject was told that if there was any trouble identifying the first supraliminal flash, it was okay to lean forward in the chair to see the subsequent flashes more clearly; however, this should not be done if the stimulus was subliminal. Then the caret test was described. Finally, the subject was asked to complete the front side of the rating scale, which asked about the subject's current mental state and expectations toward the experiment.
J. P. then went to Room 2, and A. E. sat in the hallway outside that room. The door from Room 2 to the hallway was closed. J. P. instructed the computer to print, on the printer in the hallway, whether the target slides for the subject were to have the monster in the background (monster condition) or not (no-monster condition). A. E. retrieved this information and took it to Room 3, where she turned on the projector, inserted the carousel with the proper target set, and waited for the first trial.
As soon as the subject signalled completion of the rating scale, J. P. dimmed the lights and asked the subject to put on the pair of head-phones draped over the left armrest. During the test, the subject would be listening to baroque music through the headphones.(3) A few bars of the music were played so that the subject could have the volume adjusted to comfort. Then 10 warm-up trials were completed; the procedure for these was the same as in Session 1, except that this time the subject listened to the music.
Prior to the first formal run, J. P. briefly reviewed the procedure with the subject. He then left the room during the time the message slide was being flashed, so that he could remain blind as to whether the presentation was subliminal or supraliminal. Following the flashes, the computer advanced the carousel to replace the message slide with the caret test slide. A bell sound from the computer signalled J. P. to return to Room 2. The first run then began. There was a 3-min break between the first and second runs, during which time J. P. saved the Run 1 data on the disk, turned over the music tape, and reset the carousel to load the message slide in the projector. The procedure for Run 1 was then duplicated precisely for Run 2.
For each trial, the target slide was flashed on the screen to A. E. for 100 ms, after which she recorded the target on a standard ESP record sheet. As an exploratory procedure, she also placed a check mark next to any trial on which she thought her sending was particularly effective. (This procedure produced no interesting results and will not be discussed further.)
After the second run, J. P. brightened the lights in Room 1 and requested the subject to complete the back side of the rating scale, which asked for reactions to various aspects of the procedure. He then printed out the results of the session, which also revealed the subliminality condition for the message slide. A. E. and J. P. then met together with the subject to go over the results and answer any questions the subject might have. J. P. asked the subject if he or she had become aware of the identity of the message on the slide and, if so, to state what it was. No subjects in the subliminal condition and all subjects in the supraliminal condition were able to do this. If the subject was in the subliminal condition, J. P. then revealed the message. Finally, he thanked the subject for participating and asked that he or she not discuss details of the experiment with any possible future subject.
The mean of the 48 ESP trials for both runs combined was 11.98 (SD = 3.55), which is very close to the mean chance expectation (MCE) of 12, t(39) = 0.25, n.s. The variance around MCE was 12.33, which is not significantly different from the MCE of 9 (CR = 1.58). The mean number of hits for the first run was 6.08 (SD = 2.09) and, for the second run, was 5.90 (SD = 2.55). Neither value approaches significance.
The cell means for ESP hits for the two runs in each experimental condition are listed in Table 1. The only one of these eight means to reach statistical significance was for the first run in the subliminal/no-monster condition, M = 7.80, SD = 1.81; t(9) = 3.14, p = .012.(4) Although this finding was not predicted in advance, this cell was the most likely of the eight to yield psi-hitting, according to the rationale of the study: (1) the subliminal message was expected to be more effective than the supraliminal message in evoking psi; (2) the monster was expected to produce psi-missing or tight variance, and (3) psi was considered more likely to be present in the first than in the second run because of the decline effect.
TABLE 1 MEANS AND VARIANCES(a) OF ESP HITS: SERIES I Run 1 Run 2 Total Monster Yes No Yes No Yes No Subliminal M 5.50 7.80(*) 6.30 5.80 11.80 13.60 V 4.10 6.20 8.90 5.00 13.60 13.40 Control M 5.70 5.30 5.70 5.80 11.40 11.10 V 4.10 2.70 2.90 8.60 13.00 9.30 Note. n = 10 for all cells. a Variances, which represent the deviations of hits around MCE, are listed without square-root transform. For single runs, MCE = 4.50; for combined runs, MCE = 9.00. * p = .012
The cell variances around MCE are also included in Table 1. None are significant by a two-tailed test. The cell most comparable methodologically to the Palmer and Johnson (1991) experiment was the subliminal/monster condition. The variance for this cell was 13.60, which is in the high direction and is the opposite of the Palmer and Johnson experiment.
Three-way between-within analyses of variance (ANOVA) with subliminality, monster, and run as the independent variables yielded no significant results for either hits or variance scores.
In the Palmer and Johnson (1991) experiment, counterbias responses were defined as those chosen 6 times or less in the total of 40 trials, or 15%. In the present experiment, there were 48 total trials, so that counterbias responses were defined as those chosen 7 times or less, because 15% of 48 is 7.2. Of the 40 subjects, 14 contributed a total of 75 counterbias responses. Of these, 19 were hits, with a percentage of 25.3. This result is nonsignificant; the corrected CR = 0.00. Thus there was no response-bias effect in Series I.
Pearson and Spearman correlation matrices were generated to examine relationships between (a) the total personality scores and responses on individual rating-scale items (as predictor variables); and (b) total hits and variance scores (as criterion variables). These analyses were performed on each run separately, and on the two runs combined, for each level of the subliminality and monster conditions plus the entire sample. The variance scores were normalized by taking the square root of the absolute value of the deviation between the hit total and MCE. It was decided in advance to choose the Spearman correlations whenever either the skewness or kurtosis of either variable in a correlation exceeded a z of [+ or -]1.50. Sex differences on the ESP measures were explored using t tests.
Matrices of this size will produce a number of significant results even if only chance is operating. Therefore, some judgment was exercised in deciding which significant correlations to take seriously and report below. These decisions were based on (a) the relevance of the relationship to the rationale of the study, and (b) its strength. Even the chosen relationships, however, can only be considered suggestive at this stage of the process.
Dissociation and anxiety. The mean on the 26-item dissociation scale (QED) was 11.63 (SD = 4.32). The mean on the anxiety scale (STAI), which has a range of 1 to 80, was 37.30 (SD = 7.37). The correlation between the two scales was positive but not quite significant, [r.sub.s](38) = +.30, p = .068.(5) The one ESP finding involving these predictors that seemed interesting was a positive correlation between the QED and the variance scores for Run 1, [r.sub.s](38) = +.32, p = .046. This finding meant that there was a relatively large spread in ESP hit totals among high-dissociation subjects, which raised the possibility that some other individual-differences variable might have interacted with the QED to affect ESP scores. To assess this possibility, separate correlation matrices for subjects above and below the mean on the QED were generated. This exercise revealed a significant positive correlation for high-QED subjects between ESP hits on Run 1 (ESP) and the STAI, [r.sub.p](20) = +.52, p = .014. In other words, among subjects high on dissociation, those high on anxiety were scoring more positively on the ESP task than were those who were scoring low on anxiety.
The fact that the STAI-ESP relationship was restricted to high-dissociation subjects is consistent with the rationale of the experiment because it was expected that a capacity to dissociate would be especially psi-conducive in this type of psi task. The anxiety variable seemed to affect the direction of the scoring. If the STAI is interpreted as a sensitizer-repressor scale, it would appear that high-anxiety subjects were more sensitized to the targets than were low-anxiety subjects, some of whom might have repressed the targets. Research has shown that samples of low scorers on anxiety scales such as the STAI include repressors but also persons who are just low on anxiety (Weinberger, 1990).
The purpose of the monster in the experimental design was to increase the salience of the targets. Given the preceding results with the STAI, it could be argued that the monster might facilitate psi-hitting among sensitizers and facilitate psi-missing among repressors. It was therefore decided to examine the STAI-ESP relationship separately for subjects in the monster and no-monster conditions. Still restricting the examination to high-dissociation subjects, we found that the STAI-ESP relationship was independently significant in the monster condition, [r.sub.p](7) = +.77, p = .016. It was positive but nonsignificant in the no-monster condition, [r.sub.p](11) = +.43.
Dissociative Method Scale (DMS). The rating-scale item most directly relevant to the experimental rationale is the one that asked subjects to describe their approach to taking the caret test. This item asked subjects to rank how frequently they (1) deliberately chose where to look on the screen when the caret test slide flashed, (2) simply let their gaze fall wherever it fell (random responding), or (3) felt drawn to a particular part of the screen. These responses were converted to a 6-point scale representing a measure of the extent to which subjects responded dissociatively to the caret test. The highest score (5) was obtained when the subject ranked the alternatives listed above in the order 3, 2, 1; that is, 3 ranked first. The lowest score (0) corresponded to the ordering 1, 2, 3.
The rationale for this scaling is as follows: A deliberate choice to look at a particular point on the screen indicates maximum involvement of one's normal conscious decision-making apparatus. If a subject rejects this option and instead describes the experience as having been drawn to a particular part of the screen, the implication is that the decision was made by a secondary cognitive apparatus functioning separately from the primary one. The random-choice alternative would seem to involve minimum involvement from either cognitive apparatus and, thus, to be intermediate on the continuum.
Examination of the correlation matrices revealed a positive correlation between the DMS and ESP in the subliminal-message condition, [r.sub.p](18) = +.64, p = .003. In other words, in this condition the subjects who responded most dissociatively during the caret test got the highest ESP scores in Run 1.
If the DMS indeed measures a dissociative approach to the caret test, it should correlate positively with the QED. It did, but to a disturbingly low degree, [r.sub.s](38) = +.09, n.s. This outcome raises questions about the construct validity of the DMS, a point that will be pursued later.
The results from Series I were sufficiently encouraging to justify proceeding with Series II. This series provided a strict replication of the subliminal condition of Series I and also provided the second relevant control condition: subliminal presentation of a neutral stimulus.
The significant post hoc findings from Series I became hypotheses for Series II. They were:
1. There should be significant psi-hitting in the subliminal-message (ASHLEY AND I ARE ONE)/no-monster condition for Run 1.
2. Among high-dissociation subjects (those scoring above the Series I mean on the QED) in the monster condition, there should be a significant positive correlation between anxiety (STAI) scores and ESP hits for Run 1.
3. In the subliminal-message condition, there should be a significant positive correlation between a dissociative approach to the caret test (DMS) and ESP hits for Run 1.
It was decided in advance to use one-tailed tests to evaluate these hypotheses. A significant result (p [is less than] .05) in the predicted direction would constitute a clear confirmation of the hypothesis. A nonsignificant trend in the predicted direction that was strong enough to produce significance (two-tailed) for the two series pooled would have the same evidentiary status as the significant findings from Series I; that is, a suggestive result in need of still further confirmation. Any other outcome would constitute a disconfirmation of the hypothesis, resulting in its rejection. The reader, of course, is free to apply his or her own labels to these outcomes.
Forty new subjects were selected by the same methods used in Series I, except that this sample did not include personal friends of A. E. However, this variable did not appear to have any influence on the results of Series I. The new sample contained 24 females and 16 males.
Design and Procedure
The design and procedure for Series II were identical to those of Series I, except that the supraliminal stimulus condition was replaced by a condition in which a meaningless stimulus was flashed to the subject subliminally. This slide had the same structure as ASHLEY AND I ARE ONE, except that each letter was replaced by an X.
The mean for the 48 ESP trials (ESP[T]) in Series II was 11.48 (SD = 2.73), which is very close to the MCE of 12, t(39) = 1.22, n.s. The mean number of hits for Run 1 (ESP) was 6.18 (SD = 2.29) and nonsignificant. The mean for Run 2 (ESP) was 5.30 (SD = 1.67), which is significantly below chance, t(39) = 2.65, p = .012. The variance around MCE for ESP[T] was 7.53, which does not differ significantly from the MCE of 9 (CR = 0.71).
The cell means for the two runs in each experimental condition are listed in Table 2. None of them differed significantly from chance, and the three-way ANOVA yielded no significant effects. The same is true for the variance around MCE, the values of which are also listed in Table 2.
TABLE 2 MEANS AND VARIANCES(a) OF ESP HITS: SERIES II Run 1 Run 2 Total Monster Yes No Yes No Yes No Subliminal M 6.40 5.90 5.50 5.20 10.90 11.10 V 6.80 4.10 2.90 4.40 10.70 5.30 Control M 6.40 6.00 5.30 5.20 11.70 11.20 V 5.20 4.40 3.10 2.40 5.50 8.60 Note. n = 10 for all cells. a Variances, which represent the deviations of hits around MCE, are listed without square-root transform. For single runs, MCE = 4.50; for combined runs, MCE = 9.00.
Pooling the results of the two series yielded no significant cell effects for either ESP hits or variance. Independent-group t tests comparing the supraliminal and subliminal control conditions (hereafter simply called the control conditions) yielded no significant differences for either hits or variance for the two runs individually or combined. The two control conditions were thus considered to be equivalent for the purpose of computing four-way ANOVAs, in which Series was added as a factor to the three-way ANOVAs computed for each series separately. None of the effects in the four-way ANOVAs for either hits or variance scores were significant.
Seventeen of the 48 subjects contributed counterbias responses (seven or fewer calls of a given symbol on the two runs combined). Of the 111 counterbias trials, 24 were hits (21.6%). This value does not differ significantly from chance (corrected CR = -0.71), so there was no response-bias effect in Series II.
Tests of Hypotheses and Complementary Analyses
Treatment effect. The mean ESP score in the subliminal-message/no-monster condition of Series II was 5.90 (SD = 2.13), which is nonsignificant and on the opposite side of MCE from that predicted, t(9) = 0.15. For the two series combined, this mean was 6.85 (SD = 2.16), which still does not differ significantly from MCE, t(19) = 1.76. Thus, Hypothesis 1 is rejected.
Dissociation and anxiety. Among high-dissociation (QED) subjects in the monster condition, the correlation between anxiety (STAI) and ESP  scores was in the predicted direction, but not quite significant, [r.sub.s](14) = +.37, p = .082, one-tailed. For the two series combined, however, the correlation is significant, [r.sub.s](23) = +.47, p = .017. The outcome is also significant when the correlations from the two series are converted to zs and combined by the Stouffer method, z = 2.69, p = .007. Nonetheless, because of the failure to significantly replicate Series I, Hypothesis 2 received only suggestive support.
In Series II, the QED-STAI correlation was significant, [r.sub.s](38) = +.42, p = .007. The correlation likewise was significant for the two series combined, [r.sub.s](78) = +.46, p = .002. Thus, subjects who were high on dissociation also tended to be high on anxiety. A curious finding is that the means on both scales were significantly higher in Series II than in Series I. For the QED, the mean was 11.63 (SD = 4.32) in Series I and 14.00 (SD = 4.89) in Series II (p = .010 by Mann-Whitney U test). For the STAI, the mean was 37.30 (SD = 7.37) in Series I and 41.60 (SD = 8.75) in Series II, the U significant at p = .024. These mean differences explain why the QED-STAI correlation is larger for the combined series than for either series separately.
Dissociative Method Scale. In the subliminal-message condition of Series II, the correlation between the DMS and ESP was in the predicted direction but not significant, [r.sub.p](18) = +.11. The correlation was significant, however, for Series I and II combined, albeit barely. Specifically, [r.sub.p](38) = +.31, p = .049. The combined outcome is a little more respectable by the secondary Stouffer method, z = 2.45, p = .015. Nonetheless, because of the failure to significantly confirm Series I, Hypothesis 3 received only suggestive support.
Unlike the QED and STAI, DMS scores in Series II (M = 2.80, SD = 1.43) had similar characteristics to those in Series I (M = 2.63, SD = 0.99). But in contrast to Series I, where the DMS-QED correlation was only +.09, in Series II it was strong and significant, [r.sub.p](38) = +.52, p = .001. The two correlations differ significantly from each other, z = 2.07, p = .039. If the data from the two series are nonetheless pooled, the resulting correlation is [r.sub.p](78) = +.34, p = .002. Combining the results by the Stouffer method yields z = 2.73, p = .006.
The significant DMS-QED correlation for the combined series seems to support the construct validity of the DMS as a measure of dissociative responding to the caret test; high-dissociation subjects were the ones who responded dissociatively on the caret test and also got the highest ESP scores. There is one problem with this conclusion: The DMS-ESP  relationship is attributable primarily to Series I and the DMS-QED correlation is attributable primarily to Series II. In other words, the DMS-ESP effect is strongest where the construct validity of the DMS is weakest! Is there some way to reconcile these findings?
First, a scatter plot of the DMS-QED correlation for Series I, where the linear correlation was near zero, was examined to see if there was evidence of curvilinearity. There seemed to be some indication of this, but it soon became apparent that to see the curvilinearity clearly it would be necessary to break the DMS down into categories. The obvious way to do this was to divide it into three categories according to which of the three alternatives subjects ranked first on the DMS item: Those whose first choice was deliberately choosing where on the screen to look became Group 1 (DMS-1); those who primarily focused randomly became Group 2 (DMS-2); and those who primarily felt drawn to a particular part of the screen became Group 3 (DMS-3); the higher the group number, the more dissociative the response method.
In Series II, where the DMS-QED correlation was +.52, the trend of QED scores across the three DMS categories is nicely linear. However, in Series I, where the correlation was +.09, the trend is curvilinear in form. The DMS by Series interaction on QED as the dependent variable approaches significance, F(2,74) = 2.66, p = .077.
Part of the relationship between the two curves is due to the different QED means in the two series. This factor was partialled out by lowering each of the QED means of Series II an amount equal to the difference in the grand QED means between the two series. The result, depicted in Figure 3, indicates that the difference between the DMS-QED relationship in the two series is attributable primarily to the DMS-1 group, those subjects who said that for the most part they chose deliberately where to fixate on the caret slide. The subjects who chose this option in Series I had a higher mean QED score than those who chose it in Series II. It could be that for unknown reasons some high-dissociation subjects chose a low-dissociation method of responding to the caret slide, or they responded dissociatively to the caret test and interpreted Alternative 1 on the DMS as representing such an approach.
Whatever the explanation, the above pattern is strong enough to bring into question Alternative 1 on the DMS as a valid representation of a low-dissociation response strategy. Thus, in order to assure the construct validity of the DMS, it was decided to implement what would otherwise be an inappropriate procedure: removal of the DMS-1 subjects from the sample.
Next, the DMS-ESP  relationship was evaluated in the same manner as the DMS-QED relationship. These analyses were restricted to the subliminal-message condition, because only in that condition had the relationship between the DMS and ESP been significant. When the continuous DMS scores were transformed into the three categories (including DMS-1) and examined in relation to ESP, a mirror image of the DMS-QED pattern illustrated in Figure 3 appeared. In Series I, where the correlation was strong (.64), the relationship is clearly linear. In Series II, where the correlation was weak (.11), the relationship is curvilinear in form. Again, it is the DMS-1 group that is responsible for the interseries difference.
Removal of the DMS-1 subjects had the effect of bringing the results of the two series into much closer alignment. This result can be seen most clearly by examining the relevant correlations, which are presented in Table 3.
TABULAR DATA OMITTED
When the two series were combined, the mean of the DMS-3 subjects in the subliminal-message condition (n = 7) was 7.86 (SD = 2.19), which approaches significant psi-hitting, t(6) = 2.24, p = .066. The mean of the 26 DMS-9 subjects in this condition was close to chance (M = 6.15, SD = 2.24). The difference between the two means approaches significance, t(6) = 2.24, p = .083. This analysis reveals that the deviation from MCE is attributable to the DMS-3 subjects.
Further inspection of the data for the combined series led to further refinements of the preceding preliminary conclusion. In particular, it was noted that there was a general tendency in the control conditions for DMS-3 subjects to score lower on the ESP task than DMS-2 subjects did, a reversal of the trend found in the subliminal condition. The effect in the control conditions also was somewhat stronger in Run 2, where it approached significance, t(35) = -1.83, p = .078.
To assess this broader effect, a between-within ANOVA was performed, with Series, Subliminality, DMS (minus DMS-1), and Runs as the independent variables. The analysis is presented in Table 4. A significant main effect for Subliminality is superseded by a significant Subliminality-by-DMS interaction, F(1,62) = 6.51, p = .013. Because none of the interactions involving Runs approach significance, this interaction applies to the combined ESP scores for Runs 1 and 2.
This interaction is illustrated in Figure 5. In the subliminal-message condition, the mean ESP score for the 7 DMS-3 subjects was 13.71 (SD = 3.35), and the mean for the 26 DMS-2 subjects was 11.96 (SD = 3.39). In the combined control conditions, the mean ESP score for the 8 DMS-3 subjects was 9.50 (SD = 2.20), and the mean for the 29 DMS-2 subjects was 11.52 (SD = 2.89). Note that whereas both DMS-2 groups averaged very close to the MCE of 12, both DMS-3 groups deviated from it, the direction depending on the subliminality condition. The mean of 9.50 in the control conditions was significant in the psi-missing direction, t(7) = 3.49, p = .010, whereas the mean of 13.71 in the subliminal-message condition was not, t(6) = 1.35. These two means differ significantly from each other, t(13) = 2.92, p = .012.
Although ns are very small, there is no credible evidence to suggest any real difference between the independent contributions of the two control conditions to the above effect. The ESP mean for the 3 DMS-3 subjects in the supraliminal control condition was 8.67 (SD = 3.51); for the 5 subjects in the subliminal control condition it was 10.00 (SD = 1.23). What difference there was is attributable entirely to Run 2.
The interaction depicted in Figure 5 represents a refinement of the positive DMS-ESP relationship in the subliminal-message condition that constituted Hypothesis 2. The new effect is restricted to a DMS measure that now has some construct validity, as reflected in its simple positive relationship to the QED, and it is broadened to include the subliminality control conditions and both ESP runs. In my judgment, it provides the best description of what occurred in this experiment regarding the relevant variables. Inferential conclusions must await further confirmation.
Difficulty. The correlation matrices were examined for the two series combined to determine if any additional relationships of interest would manifest. The only significant relationship found for all conditions pooled was between ESP[T] and the rating-scale item that asked subjects how difficult they found it to pick out a single caret when the caret slide flashed (DIFCU). The correlation was rs(38) = +.29, p = .009. The effect is significant in Series I, [r.sub.s](38) = +.40, p = .010, but not in Series II, [r.sub.s](37) = +.17.(6) It is significant when the series outcomes are combined by the Stouffer method (z = 2.55, p = .011). This effect is singled out for attention here both because of its apparent simplicity and because the DIFCU item had yielded significant effects in other caret test experiments (Palmer, in press).
TABLE 4 ANOVA FOR ESP SCORES AS A FUNCTION OF DMS (CATEGORIES 2 AND 3), SUBLIMINAL MESSAGE, SERIES, AND RUN Source SS DF MS F DMS(D) 0.22 1 0.22 0.05 Message (M) 37.26 1 37.26 8.07(*) Series(S) 3.98 1 3.98 0.86 D x M 30.09 1 30.09 6.51(**) D x S 0.02 1 0.02 0.004 M x S 16.91 1 16.91 3.66 D x M x S 1.15 1 1.15 0.25 Error (b) 286.34 62 4.62 Runs (R) 17.42 1 17.42 3.82 R x D 9.90 1 9.90 2.17 R x M 1.06 1 1.06 0.23 R x S 0.06 1 0.06 0.01 R x D x M 0.64 1 0.64 0.14 R x D x S 2.50 1 2.50 0.55 R x M x S 0.64 1 0.64 0.14 R x D x M x S 6.91 1 6.91 1.51 Error (w) 282.93 62 4.56 * p= .013 ** p = .006
It is also noteworthy that the DIFCU mean of 39.83 (SD = 24.18) in Series II is much higher than the mean of 25.80 (SD = 20.07) in Series I. The difference between them is significant, t(77) = -2.81, p = .006).
When the DIFCU-ESP [T] correlation was examined in relation to the original DMS, it was found that removing the DMS-3 subjects raised the correlation from [r.sub.s](77): +.29 to [r.sub.s](62): +.44, p = .0003. Among the DMS-3 subjects, the correlation was nonsignificantly negative, [r.sub.s](13) = -.36. The two correlations are significantly different, z = 2.66, p = .008. Thus, the positive DIFCU-ESP [T] correlation applies only to DMS-1 and DMS-2 subjects.
The correlation between DIFCU and DMS was positive but not significant, [r.sub.s](77) = +.20. The curvilinear relationship previously uncovered between QED and DMS suggested that a comparable analysis involving DMS and DIFCU might be profitable. Subjects once again were classified according to the primary method claimed on the DMS, and the DIFCU ratings were submitted to a one-way ANOVA. The result was highly significant, F(2,76) = 10.04, p = .0003. The easiest method was to let the eyes fall on the screen randomly (DMS-2). The DIFCU mean for these 55 subjects was 26.55 (SD = 20.99). Surprisingly, the most difficult method was the most dissociative one, DMS-3. The mean for these 15 subjects was 53.63 (SD = 17.01). For the 9 DMS-1 subjects, the mean was 35.67 (SD = 25.58).
Comprehensive analysis of the data from Series I yielded three statistically significant effects that made theoretical sense in terms of the rationale for the experiment and served as hypotheses for Series II. All involved ESP hit scores on Run 1 (ESP) as the dependent variable. The effects were: (1) significant psi-hitting in the subliminal-message/no-monster condition; (2) a significant positive correlation between ESP and anxiety (STAI) scores among high-dissociation (QED) subjects in the monster condition; and (3) a significant correlation between ESP and the use of a dissociative approach to the caret test (DMS) in the subliminal-message condition.
None of the three hypotheses received statistically significant support in Series II, although the outcome for Hypothesis 2 approached significance (p = .082, one-tailed). Hypotheses 2 and 3 were said to have received suggestive support because the relevant correlations were statistically significant for the pooled series.
Further exploratory analyses yielded one additional significant effect worthy of note: a positive correlation between ESP[T] and subjects' reported difficulty in picking out a single caret for responding on the caret test (DIFCU). The effect was restricted to DMS-1 and DMS-2 subjects.
Series I versus Series II
There seemed to be less overall psi effect in Series II than in Series I. The one tangible indication of such a dropoff is that, when the correlation matrices from the two studies were compared using standardized and predefined criteria, there were only two significant relationships between ESP scores and predictor variables in Series II compared to eight in Series I.(7) But this was not the only difference between the two series. It was assumed at the outset that, with the exception of the control conditions for the subliminal message, Series I and II would be equivalent. Three significant mean differences between the two series on predictor variables that were directly or indirectly related to the ESP scores raise questions about this assumption. The means on both the QED and STAI were significantly higher in Series II than in Series I, and more subjects reported difficulty in picking out a caret response in Series II than in Series I. These three predictors tended to be intercorrelated: In the combined series, the STAI was significantly related to both the QED, [r.sub.s](78) = +.46, p = .0002, and DIFCU, [r.sub.s](77) = +.32, p = .004; the QED-DIFCU correlation was also positive, although not quite significant, [r.sub.s](77) = +.21, p = .063. Thus, there may be a single factor operating here.
Two other close-to-significant mean differences also deserve mention as possibly contributing to this pattern: Series II subjects were more troubled that a particular direction appeared salient to them on too many trials than were Series I subjects, [M.sub.II] = 53.94, [M.sub.I] = 43.40, p = .063, by U test. Also they felt less confident at the end of the session of having demonstrated ESP than did Series I subjects, [M.sub.I] = 35.28, [M.sub.II] = 27.40, p = .078 by U test.
These trends suggest that Series II subjects had a tougher time of it in the experiment than did Series I subjects, and this could account for less psi in Series II. But this hypothesis merely begs the question of why the two series may have differed in this respect. Although there were more friends of the agent in Series I than in Series II, this variable had no significant effect within Series I on either the ESP scores or any of the predictor variables that differed between the series. As far as could be ascertained, experimenter behavior in the two series was the same, as was level of enthusiasm, and so forth. I am at a loss to explain the interseries differences.
Dissociation and anxiety. A positive correlation was found between trait anxiety and ESP hits among high-dissociation subjects in the monster condition. The QED was included because it was felt that subjects with dissociative tendencies would be especially prone to manifest psi with a test procedure that encourages dissociative tendencies and indeed was designed to make dissociative mental responses the basis of the behavioral responses. The monster was included in the design because it was felt that it would serve as an anxiety-evoking stimulus if picked up psychically by the percipient, even unconsciously.
It is not entirely clear on a priori grounds what prediction is proper for the direction of the anxiety-ESP relationship under these circumstances. As a general rule, measures of neuroticism correlate negatively with ESP scores; low-neurotic subjects score more positively than high-neurotic subjects (Palmer, 1977). However, neuroticism embodies two somewhat contradictory concepts. On the one hand, it incorporates the notion of psychological defense mechanisms, the denial, repression, or even transformation of anxiety-arousing stimuli. The Defense Mechanism Test (Kragh & Smith, 1970) is a relatively pure representation of this aspect of neuroticism, and it correlates negatively with ESP scores quite consistently (Haraldsson, Houtkooper, & Hoeltje, 1987; Watt, 1991). Pure anxiety, on the other hand, is an absence of defense mechanisms; instead of its being defended against, the anxiety is fully experienced by the person, who in effect is sensitized to anxiety-arousing stimuli. Such a person may try to avoid anxiety-evoking situations behaviorally, but when confronted with them, he or she experiences them realistically.
This line of thought would lead to the expectation that high-anxiety subjects should score high on ESP in the monster condition, because the monster would sensitize subjects to the targets. In fact, Palmer (1977) found that the "neuroticism" measure that has resulted most frequently in positive correlations with ESP scores is the Taylor Manifest Anxiety Scale (MAS), which correlates .80 with the STAI (Spielberger, 1983). However, the MAS has also yielded some negative anxiety-ESP relationships, and one in particular is relevant to the present experiment. Using high-school students as subjects in a clairvoyance task, Carpenter (1971) found a negative relationship between ESP scores and the MAS when erotic photographs were secretly included with the ESP target symbols, but a positive relationship when the erotic photos were excluded. If the erotic photos can be considered as functionally equivalent to the monster in the present experiment, then the expectation should be a negative STAI-ESP correlation in the monster condition.
The finding of a positive STAI-ESP correlation in this condition can be reconciled with Carpenter's finding if one supposes that the monster was threatening enough to capture the attention of high-dissociation, high-anxiety subjects but not as threatening as the erotic photos Carpenter used. In short, the hypothesis becomes that a mildly threatening stimulus would sensitize high-anxiety subjects to itself, whereas a strongly threatening stimulus would overwhelm them and cause avoidance.
Dissociative Method Scale. Post hoc analyses of data from the pooled series led to a modified description of the positive DMS-ESP  relationship in the subliminal-message condition. An interaction that was uncovered indicated that the direction of scoring among subjects who frequently felt drawn to a particular part of the screen when the caret slide flashed (DMS-3) was significantly influenced by the subliminality manipulation: DMS-3 subjects who received the message ASHLEY AND I ARE ONE subliminally scored above chance, whereas those who received either this message supraliminally or a meaningless stimulus subliminally scored below chance. The above-chance effect was stronger in Run 1, and the below-chance effect was stronger in Run 2. This latter outcome is a consequence of generally more negative scoring in Run 2 than in Run 1; in Series II, the overall mean of Run 2 was significantly below chance.
The interaction just described is consistent with the thesis of Silverman that certain subliminal messages can have beneficial effects on task performance, although it should be kept in mind that the mean in the subliminal-message condition was not significantly above chance, even for DMS-3 subjects. The subliminality of the message was nonetheless shown to be important because supraliminal presentation of the message had the opposite effect. It is puzzling that the control stimuli led to psi-missing (especially in Run 2) rather than strictly chance performance. Perhaps control subjects realized, consciously in Series I and unconsciously in Series II, that they were indeed control subjects and this caused a rebellion of sorts. Conscious recognition of the ASHLEY AND I ARE ONE message might also be expected to be a little disconcerting to some subjects. It is also conceivable that the Xs used in the subliminal control condition had a negative connotation for some subjects (e.g., an X is often used to indicate wrong responses on a test). However, this possibility strikes me as a bit far-fetched.
The most intriguing aspect of this finding is that the effect of the subliminality manipulation seemed to be restricted to those subjects who responded dissociatively to the caret test. It is reasonable to suppose that if these DMS-3 subjects were in a mildly dissociative state during the caret test runs, they were also in such a state during the prerun periods when the message stimuli were flashed. If they were, this would suggest that subjects are more receptive to subliminal influence in a dissociative state. Extrapolating a bit further, this finding might suggest that hypnosis would facilitate subliminal receptivity. This speculation, of course, would need to be tested directly.
Difficulty. The significant positive correlation between the DIFCU item and ESP[T] confirmed the results in the DMT-ESP experiment of Palmer (1992), but not the results in the other caret test experiment, which used a different caret slide (Palmer, in press). One of the ideas behind the perceptual ESP test was to make it easier for subjects to choose responses than it is in conventional ESP tests. But not only did some subjects rate this process as somewhat difficult in the caret test, on the whole these subjects got the highest ESP scores. Perhaps correct paranormal information somehow interacts with the information conveyed by the strictly perceptual response, making the choice harder when such paranormal information is present.
But why did the positive DIFCU-ESP[T] correlation reverse for DMS-3 subjects? A clue might be hidden in the fact that among DMS-3 subjects, the DIFCU mean was suggestively higher in the control conditions (M = 60.00), where there was significant psi-missing, than in the subliminal-message condition (M = 46.35). (The mean for the remaining subjects DMS-1 + DMS-2 was 27.83.) Perhaps a certain amount of difficulty is conducive to positive scoring, but when the difficulty is too great it becomes detrimental. We might speculate that for some reason the control messages adversely affected those DMS-3 subjects who were exposed to them, such that they experienced the caret task as more difficult and therefore scored below chance on it.
An exhaustive analysis of data from 80 subjects in two experimental series led to the identification of several statistically significant relationships involving ESP hit scores. Most of these effects make sense in terms of the theoretical rationale of the project, although in some cases they have caused this rationale to be modified or at least refined.
The extensive dissection of these effects, although tedious, was necessary if they were to be interpreted in the way that is truest to all the data. In some cases, these interpretations were different than they would have been had the complementary analyses not been undertaken, particularly with regard to the boundary conditions of the effects.
These findings are valuable primarily as guideposts for future research. It is not proper to claim more than this because none of them have been significantly replicated. It is worth noting in this connection that many of these relationships were confined to subgroups of the samples, so statistical power was often low. If these effects are genuine, larger samples may be needed to reliably detect them.
1 The blinds conceivably could have been compromised for the last couple of subjects, because the total number of subjects in each condition was fixed. However, neither of the experimenters was keeping count.
2 During Session 2, the message slide was flashed at 1 ms for this subject.
3 The tape was entitled, "Pachelbel Canon: Handel, Vivaldi, Gluck," by the Academy of Ancient Music and Christopher Hogwood (KDSLC 594).
4 All p values are two-tailed unless otherwise noted.
5 Both of these scales include items that transparently ask about behaviors, cognitions, or feelings that could be construed as socially undesirable. Thus, both are potentially vulnerable to demand characteristics. Relevant data on this point are available from 38 subjects who participated in a yet-to-be completed replication of the Palmer (1992) experiment, which, from the subject's point of view, was procedurally very similar to the present experiment. Subjects in this experiment completed both the QED and the Marlowe-Crowne Social Desirability (MCSD) Scale (Crowne & Marlowe, 1960). Although the MCSD scale is now interpreted primarily as a measure of psychological defenses (Weinberger, 1990), one would expect subjects who are susceptible to demand characteristics on personality tests to score high on it. The correlation between the QED and the MCSD scale was weak and nonsignificant, [r.sub.p](36) = -.08, suggesting that demand characteristics were not a problem, at least not with the QED.
6 one subject in Series II did not complete the item.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||includes questionnaire|
|Publication:||The Journal of Parapsychology|
|Date:||Jun 1, 1994|
|Previous Article:||Psyche and Matter.|
|Next Article:||Anomalies of consciousness: Indian perspectives and research.|