Experimenter effects and the remote facilitation of attention focusing: Two studies and the discovery of an artifact.
Some researchers have noted that certain parapsychologists have a tendency to get positive psi results whereas others tend to obtain chance results (e.g., Kennedy & Taddonio, 1976; Palmer, 1997). Several different mechanisms have been proposed to account for this effect. For example, experimenter personality and other physical characteristics may vary so that participants will respond differently to different experimenters. Some experimenters may be better at eliciting psi from their participants than others, by creating a more psi-conducive atmosphere (e.g., Honorton, Ramsey, & Cabibbo, 1975; Schneider, Binder, & Walach, 2000). In addition, if the participant believes in the possibility of psi, perhaps they are more likely than the disbelieving participant to respond to positive suggestions of success from the experimenter. It is also possible that experimenter effects may reflect experimenter psi. Also, some experimenters may use different methods, some of which are more psi-conducive than others. lastly, some experimenters may be more adept than others at selecting participants likely to be able to demonstrate psi in experiments (Parker, Frederiksen, & Johansson, 1997).
Clearly, the question of experimenter effects in parapsychology is a complex one. Careful research is needed to untangle these possible contributory factors. One notable attempt to do this is with the collaborative studies of skeptic Richard Wiseman and psi proponent Marilyn Schlitz. Each had conducted studies into remote detection of staring, using electrodermal activity (EDA) as the dependent variable. Wiseman had obtained null results, whereas Schlitz had obtained positive results. Aware that there might be differences in how they conducted their experiments and in the kind of participants they recruited that might contribute to this apparent experimenter effect, Wiseman and Schlitz conducted a series of joint studies in which they used the same experimental equipment and procedures and the same pool of participants, thus controlling for these factors. Each acted as remote observer for their participants. Their findings continued to replicate the pattern of their earlier research, with Wiseman obtaining nu ll results and Schlitz obtaining positive results (Wiseman & Schlitz, 1997, 1999).
The experimenter effect obtained by Wiseman and Schlitz could be due to experimenter psi or differences in experimenter--participant interaction (or a combination of both), as the two differ in their personality, physical characteristics, and their prior beliefs about psi while their studies controlled for differences in participant population and methods. The two also differ in their prior reputation as having a track record of positive psi results versus a record of null psi results, and this too may have been a factor in their experimenter effect.
The present research aims to build on Wiseman and Schlitz's work by eliminating differences in experimenter/influencer variables by using a single experimenter and influencer (Caroline Watt [CW]) throughout. CW has a mixed research track record of occasional positive and more often null psi results, and she regards herself as on the positive shade of neutral in attitude towards psi. The studies aim to restrict any possible experimenter effect to one factor: the participants' prior perception of the psi research "track record" of their experimenter. This was achieved by asking participants to read a short article that either portrayed CW as having a previous history of positive psi results (positive expectancy condition) or as having a track record of null psi results (negative expectancy condition). This attempt to influence participants' expectancies about their experimenter's previous psi research track record is similar to several previous parapsychological studies in which participants' expectancies and a ttitudes have been varied. For instance, Layton and Turnbull (1975) exposed participants to either a positive or a negative evaluation of ESP prior to ESP testing and found, for one of their two experimental series, that the positive group had higher ESP scores than the negative group. Prior to ESP testing, Taddonio (1975) gave participants written instructions that included suggestions about the past performance of participants on the task. As predicted, the group that expected to score above chance scored significantly better than the group that expected chance scoring.
In the present study, during the presession chat, CW attempted to treat the participants identically, with the exception that for the positive expectancy condition she suggested to participants that she had previously been successful with this kind of research and was optimistic for a positive outcome. This suggestion was not made in the negative expectancy condition. The aim was to simulate what probably goes on in the majority of parapsychology experiments. That is, all experimenters, even skeptical ones or those with a track record of null results, presumably attempt to be friendly and set their participants at ease. It is likely, however, that proponents or those with a history of positive results are more likely to make implicit or explicit positive suggestions for success than skeptics or those with a history of null results.
For the psi task, participants were asked to focus their attention on a candle and to indicate by pressing a button when they realized they had become distracted from their focus. At the same time, and according to a random schedule, a remote "helper," who is also the experimenter, focused on a similar candle with the intention of facilitating the attention focusing of the "helpee." At other times, the helper did not focus on the candle. This psi task has been successfully used in two previous studies (Brady & Morris, 1997; Braud, Shafer, McNeill, & Guerra, 1995) and was selected because we felt the methodology was simpler than the physiological direct mental interactions with living systems (DMILS) paradigm and because of our interest in direct mental influence on behavioral and cognitive measures. The present study replicates the psi methodology of the previous studies, having the same dependent variable for the psi task (i.e., registered distractions) and the same number, duration, and counterbalanced rand om scheduling of help and control epochs.
Questionnaire measures of participants' belief in psi and personality were obtained prior to the psi task but before any interaction with CW. The belief in psi question is of particular interest because Wiseman and Schlitz (1997, 1999) found that participants' belief was lower for Wiseman than for Schlitz, raising the possibility that the experimenters were eliciting divergent responses from their participants (the belief questionnaires were administered following participants' presession interaction with the experimenters). In the present study, we therefore examined whether participants' belief scores varied according to their expectancy condition. For additional exploratory investigation, questionnaire measures of participants' expectations of success and their evaluation of the experimenter were obtained before and after the psi task.
The 60 volunteer participants, paid [pounds sterling]5 each for taking part, were undergraduate students, mostly studying psychology. There were 18 male and 42 female participants, with a mean age of 20.4 years (SD = 3.3) and a range from 17 to 35 years.
Priming Article. This simulated one-page article from a fictitious psychology magazine was designed to manipulate participants' expectancy before completing questionnaires and participating in the study. There were two versions of the article. They were identical except that the positive expectancy version noted that "Dr Watt is one of those who has a history of positive results, saying that 'in my 12 years as a researcher I have generally got positive results in my experiments.'" The negative expectancy version noted that "Dr Watt is one of those who has a history of chance results, saying that 'in my 12 years as a researcher I have generally not got positive results in my experiments.'" For additional emphasis, the quote was repeated in an inset.
Participant Information Form (PIF). This was a 77-item questionnaire including 1 item on participants' belief that they could demonstrate psi in a controlled experiment and 11 items on belief in and experience of telepathy, clairvoyance, precognition, and psychokinesis. These terms were all defined for the participants. High scores indicated higher belief; scores could range from 12 to 84. The PIF also contained two questions on tendency to become absorbed in a task: "How often do you lose awareness of your surroundings when you get involved in an activity?" and "How often do you lose your sense of time when you get involved in an activity?" (scored on a 7-point scale ranging from 1 = never to 7 = always), and two questions on experience and practice of meditation: "Do you practice any form of mental discipline/exercise, e.g., meditation, biofeedback, hypnosis, relaxation exercises?" (scored on a 7-point scale ranging from 1 = never to 7 = daily), which might be relevant to participants' performance on the pr esent experimental task.
NEO Five-Factor Inventory Questionnaire. One reason for obtaining personality scores was to explore whether there was any interaction between personality and expectancy condition, with regard to psi scores and to responses on the session questionnaire measures. For instance, are certain personality factors associated with more responsiveness to the expectancy manipulation than others? The NEO Five-Factor Inventory is a well-established 60-item short form of the NEO personality questionnaire (Costa & McCrae, 1991) that measures the five personality factors of Neuroticism, Extraversion, Openness, Agreeableness, and Conscientiousness. For each factor, scores can range from 0 to 48. This was administered to collect additional information on possible personality correlates with the other measures being taken in the study.
Session Questionnaire. A five-item questionnaire acted as a check on the expectancy manipulation and to gather information on how expectancy effects might operate. Participants were asked to complete one item just prior to commencing the focusing task: "How confident are you that you will have fewer distractions during the periods when Caroline is attempting to help you focus, compared to the control periods?" (five response options ranging from not at all confident to very confident, with higher scores indicating greater confidence). Participants completed the remaining four items immediately after the focusing task. First, "To what extent do you feel that you were being helped to focus by Caroline?" (five response options ranging from not at all to completely, with higher scores indicating greater perceived helping). The remaining three items asked "How would you rate Caroline as an experimenter?" Response option ranges were warm--cold (Q3), professional--unprofessional (Q4), and instilling confidence for t ask-not instilling confidence for task (Q5), again each with five response options, with higher scores indicating more positive evaluations of the experimenter. During the session, the questionnaire was folded so that participants could only see Question 1 at the start of the session.
The initial chat between CW and the participant was video recorded to enable future examination of this interaction. Participants were aware of the recording and gave their consent to this. This is not reported in the present article.
The helpees were located in a windowless sound-attenuated room with two foam-padded doors, within a suite of experimental rooms entered by a third door, on the top floor of the psychology department. Helpees were seated in a comfortable chair; their room also contained a video camera, a response button, and a small table holding a wrought-iron support for a blue translucent glass candle-holder and a candle. The helper was located in a separate room 20 m away up a short flight of stairs, with a window whose blinds were drawn. The helper was also seated in a comfortable chair and with a similar table and candle-holder, and a computer monitor and headphones. The computer controlling the influence program was housed in a third closed room (the control room) adjacent to the helpee's room. The rooms were not connected by any ventilation system.
Additional Security Measures
In both studies, the helper was blind to the influence schedule until separated from the helpee. The helpee was blind to both the distribution and duration of the influence periods and did not know at what point the influence schedule began. Session data were recorded in duplicate, both to computer disk and to hard drive. The helpee did not know the location of the controlling computer. The controlling computer's monitor did not display any information about the influence schedule, and if the controlling computer was interfered with during the session, this would affect the display seen on the helper's monitor.
The helpee's task was to focus his or her attention on a lit candle and holder and press a hand-held button whenever they noticed they had become distracted. The response button was connected to the controlling computer, programmed in QuickBasic by Paul Stevens (PS). The first screen had a box for the participant's ID number and a "start" button. The next screen had the message "waiting 60 seconds." After the appointed time delay, the control monitor displayed "session in progress" while the helper's monitor began to display the influence schedule to the helper. The helper had no feedback during the session of the helpee's button presses. The schedule was arranged to replicate that in Braud et al.'s (1995) and Brady and Morris's (1997) studies. There were 16 one-minute epochs, arranged in 8 pairs: 4 help--control pairs and 4 control-help pairs. The order of presentation of the pairs was randomly arranged for each session by pseudorandom algorithm within the program, once the program was initiated. At the star t of each epoch, an auditory tone was played over the helper's headphones to remind her to check the monitor for instructions. The tone was identical for help and control epochs. During each epoch, the helper's monitor displayed either "Help" or "Control." On completion of the 16 epochs, the helper's monitor displayed "end of session."
Claire Brady (CB). contacted participants by e-mail or word-of-mouth. She gave them a brief description of the study, omitting any mention of the expectancy manipulation, and noted that CW would be their experimenter and remote helper. If they agreed to participate, they were then sent the background article and the PIF and NEO questionnaires. They were explicitly requested first to read the background article, then to complete the PIF, and finally to complete the NEO.
At the appointed time, CB met the helpee, settled them in the testing room, and gave them refreshments if desired. GB then called for CW and commenced the video recording, then left. CW collected the questionnaires from the helpee, checked that he or she had completed them correctly, and verified that the helpee had indeed read the background article. She then described the procedure, noting that the procedure was like a joint meditation at the times when she was focusing with the helpee. For those helpees in the positive expectancy condition, CW noted that she had previously done a similar study with positive results and was therefore optimistic for a good outcome. She did not say this for the negative expectancy helpees. Apart from this difference, she attempted to treat all helpees identically. She noted that helpees would get feedback on the psi task at the end of the session and that they would receive further written feedback once the entire study was completed. After answering helpees' questions (but g iving no information on the nature of the influence schedule or the expectancy manipulation), CW switched off the video camera and took the helpee to see the helper's room, lit the candle in this room, and described what she would be doing during the help and control epochs. CW stressed that during the help epochs, she would not be consciously striving to help the participant, merely attempting to focus on the candle and have the hope in the back of her mind that her focusing would help the participant to focus too. She pointed out that during the control epochs she would be doing some irrelevant reading task.
On returning to the helpee's room, and while the helpee completed the presession questionnaire, the helper lit the helpee's candle and positioned it about 2.5 m in front of them, at about eye level. After ascertaining that the helpee had no more questions and was ready to proceed, the helper wished the helpee an enjoyable session and instructed the helpee to commence the focusing task immediately after the room doors were closed. The helper switched off the lights and closed the doors of the room.
The exact time of commencement of the psi task then varied. Often CW went directly to the control room and initiated the influence schedule (which was preceded by a 60-s delay to give CW time to reach the helper's room), but occasionally the start was delayed by a further minute or two as, for example, CW made herself some refreshments. At no time during the session did CW interact with other laboratory staff.
The computer recorded the participant ID, session date and time, a record of the help--control sequence, summary statistics of the total number of button presses for each condition, and a percentage influence score (PIS) calculated from the ratio of help to control presses (sum of control divided by sum of help and control presses; Braud & Schlitz, 1991). Exact chance results would have a PIS score of 0.5; results in the predicted direction (fewer distractions in the help condition) would give a score greater than 0.5.
During the psi task, CW remained silent and seated. At the end of the psi task, CW then left her room, opened the doors to the helpee's room, and turned on the lights in that room. The helpee completed the remaining session questions. CW then briefly discussed the helpee's experiences during the session and gave the helpee [pounds sterling]5 for participation. Finally, both went to the control room to view the results. The results were not known by helper or helpee until this point. The results were displayed on the monitor in the form of number of H (help) presses, number of C (control) presses, and PIS score. CW explained the session data to the participants, who were then thanked before they left. Sessions lasted approximately 40 mm.
Design, Hypotheses, and Planned Analyses
The study was planned to end after 60 participants, 30 in each condition, had been tested. GB allocated participants approximately alternately to positive expectancy or negative expectancy conditions, while attempting to balance the number of male and female participants in each condition. There were two primary hypotheses. We predicted that participants would show fewer distractions during the help periods compared with the control periods (Hypothesis 1). To facilitate comparison with previous remote facilitation of attention studies, we used t tests (one-tailed) to compare sum of help presses with sum of control presses for each participant; we also report the effect size (Cohen's r). For Hypotheses 2 and 3, a single outcome measure was needed to summarize the remote influence effect for each session. PIS scores were used for this purpose. (1) Hypothesis 2 predicted that participants in the positive expectancy condition would have higher PIS scores than those in the negative expectancy condition. Following the experiment, participants were categorized as believers in psi or disbelievers in psi according to a median split on the questionnaire measure of belief. A secondary prediction (Hypothesis 3) was that belief would interact with the expectancy condition, such that the highest PIS scores would be for the believers in the positive expectancy group. Hypotheses 2 and 3 would be tested by 2 (believers vs. disbelievers) x 2 (positive expectancy vs. negative expectancy) analysis of variance (ANOVA). Additional exploratory questions would consider the effect of belief, personality, and expectancy condition on session questionnaire responses.
Because of a computer artifact (see below for more details), the psi results from this study could not be evaluated. Therefore the analysis is limited to the questionnaire measures of belief in psi, expectancy and perception of success, and the experimenter evaluation questions.
Belief in Psi
Mean belief score was 48.3 (SD = 14.6; range = 19-82). Mean belief scores for positive and negative expectancy participants respectively were 46.6 (SD = 15.5) and 50.0 (SD = 13.6). There is no significant difference between these belief scores (Mann--Whitney z = -.710, p = .478, two-tailed), and therefore there is no indication that participants' belief scores were influenced by their reading of the priming article prior to completing the belief questionnaire. Wiseman and Schlitz (1997, 1999) found their participants differed on belief scores. This may have been because they administered the belief questionnaire after participants had met and had an introductory chat with the experimenter, whereas the present study took the participants' belief measure before they met the experimenter. Wiseman and Schlitz's participants may therefore have been subjected to more cues as to the experimenter's belief prior to completing the belief questions, and this may have exerted a greater influence on participants' question naire responses. Also, the belief questions in Study 1 came at the end of a 77-item questionnaire, so perhaps the priming article containing the expectancy manipulation had lost some of its salience by the time participants did the belief questions. For this reason, Study 2 would present the belief items in a shorter questionnaire.
Table 1 shows participants' responses to the session questionnaire. For all questions, mean responses in the positive expectancy condition were nonsignificantly higher than in the negative expectancy condition (tested by Mann--Whitney). This could just be a chance trend. Study 2 would look at this question again to try to clarify whether participants' expectancy condition affected their session questionnaire responses.
Mean personality scores (SD in parentheses) for each of the five factors are as follows: Neuroticism, 22.6 (8.0); Extraversion, 30.9 (5.5); Openness, 33.0 (6.1); Agreeableness, 32.8 (5.9); and Conscientiousness, 28.1 (7.1). Spearman correlation coefficients were calculated between these factors and participants' responses on the session questionnaire measures, as shown in Table 2. There was a significant positive correlation between Extraversion and participants' ratings of the experimenter instilling confidence, [r.sub.s] = .337 p < .01, two-tailed. Owing to the number of analyses, this correlation could be spurious, so it is regarded as a suggestive trend until further confirmation is obtained in follow-up studies. Apart from the trend, there seems to be little relationship between personality and participants' ratings during the session.
Personality, Expectancy, and Session Ratings
For each personality factor, participants were split on the median into high and low groups. We conducted 2 (expectancy condition) x 2 (high or low personality factor) ANOVAs with responses to each of the five session questionnaire items listed in Table 2 as dependent variables. No significant interactions were found with these analyses. Therefore, the present study gave no indication that participants with differing personalities responded differently to the different expectancy condition with respect to their session questionnaire measures.
Absorption, Meditation Practice, and Psi
Because of the computer problem, it was not possible to conduct exploratory measures of the relationship between participants' tendency towards absorption, meditation practice, and the outcome of the psi task. This question would be examined again in Study 2.
Artifact in Study 1: A Cautionary Tale
We describe why we were unable to report on the psi results of Study 1, as we feel future investigators may benefit from our experience. In the final Discussion section, we consider how the discovery of the artifact relates to questions of experimenter effects.
The study results showed a large apparent psi effect in the predicted direction. Before, during, and after Study 1, we conducted a number of manual checks on the computer program, as well as randomization checks. Some of this checking was done with the help of Richard Wiseman. We were unable to identify an artifactual explanation for the apparently strong psi results. We therefore decided to replicate the study including an alteration to the program that might help identify the factors underlying the positive results of Study 1, whether these might be artifactual or possibly psi-related factors. The program in Study 2 recorded the exact time of the participant's button press. When testing began in Study 2, GW and PS discovered that if participants pressed the response button prior to commencement of the influence schedule, the computer program erroneously assigned the button press to the default value of "C" (for control). The subsequent calculation of the PIS score included these "extracurricular" button pre sses, which resulted in an artifactually inflated number of C presses and a more positive PIS score.
Although this flaw may seem obvious in hindsight, in fact it took us a lot of time and effort to detect it. There were two reasons why we did not detect this before or during Study 1. First, the program only recorded summary statistics, so it was less easy to track events during a session. Second, our pilot sessions and manual checks on the computer program did not exactly duplicate what participants were doing. We simulated participants' behavior during the 16 influence periods and found no discrepancy. However, because we the experimenters acted as helpees during the pilot sessions, and because we had a model in mind that what was important was the participants' behavior during the influence schedule, we did not simulate the occasional button press that participants would make during the time delay prior to the commencement of the influence schedule. This was an important lesson for us to learn, which of course has wider implications for any researcher. Pilot testing should be done with the same kind of par ticipants as the formal study, and using exactly the same procedure as the formal study. Otherwise, artifacts may remain undetected. Note that if we had conducted an exact replication of Study 1, this artifact could still have remained undetected, which would have led to dissemination of a spurious positive result.
Study 2 was a replication of Study 1, conducted to gather further information that might throw light on the apparent positive psi findings of Study 1 (recall that the discovery of the artifact in Study 1 did not occur until Study 2 was under way) and to reexamine the nonsignificant trends in Study 1. Study 2 contained some adjustments to the questionnaire measures to gather additional detail. The same number and type of participants, the same expectancy manipulation, the same order of administration of questionnaires and psi task, and the same planned hypotheses and analyses were used. Procedural changes, and the reasons for them, are noted below. In particular, the psi results in Study 1 did not contain the artifact that had affected Study 1.
There were 20 male and 40 female participants, with a mean age of 21.2 years (SD + 5.4) and a range from 17 to 52 years.
Priming Article and NEO. These were the same as used in Study 1.
PIF. The abbreviated form used in Study 2 simply included demographic questions and the aforementioned belief, absorption, and meditation questions (a total of 20 items). The form was abbreviated so that participants would be answering the belief questions almost immediately after reading the priming article.
Session Questionnaire. The same five items plus two new questions were used in Study 2. The first new question was administered prior to the psi task, after the question on confidence of success at the psi task. The new question varied depending on condition and was intended as a "boost" to the expectancy manipulation and to obtain participants' self-report on the expectancy manipulation: "Do you think your confidence is affected by the fact that Caroline's past experiments have generally given positive results?" (tick one of three boxes labelled Yes, No, or Unsure) (positive expectancy) or "Do you think your confidence is affected by the fact that Caroline's past experiments have generally not given positive results?" (same response options) (negative expectancy). The second new question, administered after the psi task, was intended to obtain further information on participants' distractions: "In general, what criteria were you using to decide what counted as a 'distraction'? (please tick one or more boxes below)." There were four response options: "external" distraction (with examples); fleeting "internal" distraction (with examples); longer "internal" distraction (with examples); and other (please describe). During the session, the questionnaire was folded so that participants could only see Questions 1 and 2 at the start of the session.
Although we had planned to record experimenter--participant interactions in Study 2, no video equipment was available; recording was therefore not done in Study 2.
Because of other research going on in the rooms that had been used in Study 1, Study 2 used two different rooms, again on the top floor of the psychology department. The helpee's room was a windowless sound-attenuated room, with two doors that were not foam-padded, located in an experimental suite entered by a third door. As it was a smaller room than the helpee's room in Study 1, and less well sound-proofed, two changes were introduced. First, the ventilation fan in the room was switched on during the session, for ventilation and to help mask any other background noises from outside of the room. Second, helpees were asked to wear sound-proofing headphones during the meditation task, to minimize any distracting noises from outside the room. We did not consider that there was a problem with sensory leakage between the helper's room and the helpee's room. However, we were concerned that the helpee might be distracted by other activities going on in the other rooms in the experimental suite. The helper's room wa s CW's office, located about 25 m away from the helpee's room, up a short flight of stairs. During the session, the window blinds were lowered and CW's phone calls were diverted to avoid interruptions. There was no ventilation link between the helper's and helpee's room. The helper's room also contained the control computer.
Additional Security Measures
Unknown to participants, an adhesive paper tab was placed over the outer door to the helpee's room during the period of the psi task, so the helpee could not leave the room without detection (there was no sign that this tab was ever disturbed during the study).
The instructions to the participant remained unchanged. However, as the control computer was in CW's office, only one monitor was used. After CW pressed the start button, there was a 15-s delay before the influence schedule commenced. At the end of the 16 epochs, the helper's monitor did not reveal the results until CW pressed a button saying "show results"; this she did when all other session measures had been completed. This was to ensure that CW and the helpee remained blind to the results until the end of the entire session. To gather additional information about participants' responses during the session, we changed the computer program to record the time at which participants pressed the response button, in addition to the other data recorded as before.
With the exception of those changes already noted, the procedure for Study 2 was identical to that for Study 1.
Remote Facilitation of Attention, Expectancy, and Belief in Psi
The mean number of registered distractions in help and control epochs was 12.58 and 12.20, respectively. This compares with 12.43 and 13.60 for Braud et al.'s (1995) study and 18.45 and 19.60 for Brady and Morris's (1997) study. A related t test comparing mean number of help presses with mean control presses showed a nonsignificant trend in the direction opposite to that predicted: t(59) = -0.823, ES = -.11, SD = 0.02. [Effect size formula r =[[t.sup.2] / ([[t.sup.2] + df)].sup.1/2]; Rosenthal & Rosnow, 1991.] No p value is reported because the trend is opposite to that predicted on the planned one-tailed test. There was therefore no support for the hypothesis that participants would show fewer distractions during help epochs compared with control epochs. This analysis fails to replicate prior research suggesting remote facilitation of focusing of attention (Brady & Morris, 1997; Braud et al., 1995).
The median belief score was 48. Because there were 3 participants with this score and we wanted to have equal numbers of believers and disbelievers, we decided to designate as disbelievers the 1st participant (in order of testing) scoring 48 and the other 29 participants with scores less than 48. The mean belief score for disbelievers was 37.8 (SD = 8.2). We designated as believers the other 2 participants scoring 48, along with the 28 others scoring over 48. The mean belief score for believers was 56.1 (SD = 6.1). These designations were made before any analyses were conducted.
Table 3 shows the mean number of registered distractions and PIS scores broken down by belief and expectancy condition (there were 13 believers and 17 disbelievers in the positive expectancy condition, and 17 believers and 13 disbelievers in the negative expectancy condition). ANOVA (with believers/disbelievers and positive/negative expectancy as factors and PIS as the dependent variable) found no main effect of expectancy, so there was no support for Hypothesis 2 of greater remote facilitation of attention for participants in the positive expectancy condition compared with those in the negative expectancy condition, F(1, 56) = 0.325, p = .571. The latter analysis suffers from a drop in statistical power, as there were only 30 participants in each expectancy condition, half the number of the overall analysis. The predicted interaction of belief and expectancy (Hypothesis 3) showed no significant effect, F(1, 56) = .042, p = .839. There was little correlation between belief in psi and PIS scores, [r.sub.s] = . 021.
Belief in Psi
Overall mean belief score was 46.9 (SD = 11.7). There was little difference in the belief scores for the positive expectancy condition (M = 46.6, SD = 13.1) compared with the negative expectancy condition (M = 47.3, SD = 10.5; Mann--Whitney z = -.192, p = .848, two-tailed), as was found with Study 1.
Whereas Study 1 found a consistent pattern of higher ratings for the positive expectancy condition, this was not the case for the questions on confidence of success and perceived helping in Study 2: Participants had nonsignificantly lower ratings in the positive expectancy than negative expectancy condition (see Table 4). However, as with Study 1, Study 2 also found consistently more positive experimenter evaluations in the positive expectancy condition compared with the negative expectancy condition (see Table 5).
Meditation and Absorption
Braud et al. (1995) found greater remote helping for those participants with a "need" to be helped. Brady and Morris (1997) did not confirm this finding. The present study gathered additional information that might be relevant to need for helping, with the questions on meditation and absorption. Thirty-eight participants said they never practised a mental discipline/exercise such as meditation or relaxation (the nonmeditators), whereas the remaining 22 said they did practise meditation or relaxation, even if only rarely (the meditators). Looking at the number of button presses as an indicator of distractions, the meditators registered nonsignificantly more distractions per session than the nonmeditators (meditators mean of 28.4 distractions, SD = 14.04; nonmeditators mean of 22.7 distractions, SD = 17.09), t(58) = 1.331, p = .188, two-tailed. While one might have expected the meditators to have had fewer distractions, most participants were only occasional meditators. Therefore, they may have been at the earl y stages of meditation expertise, where the first step is to notice that distractions are occurring. There was no significant difference between the two groups in their performance on the psi task (nonmeditators, mean PIS score = 0.507, SD= 0.09; meditarors, mean PIS score = 0.483, SD= 0.11), t(58) = 0.896, p= .374, two-tailed. There was no correlation between total number of button presses per session and the participant's absorption score, [r.sub.s] = -012. There was a nonsignificant positive correlation between absorption and PIS score, [r.sub.s] = .187.
It is difficult to interpret the information about absorption and practice of meditation, because they are rather indirect measures of need for helping. One might argue that participants who report becoming easily absorbed might find the focusing task easier and might therefore have less need for remote helping. The indirect measures in the present study therefore tend to align with Brady and Morris's (1997) findings on "need for helping"; however, a more direct measure of need for helping, such as directly asking participants to rate how easy they had found the focusing task, might be more informative for the question of the relationship between need for helping and PIS.
The mean Neuroticism score was 22.7 (SD = 7.4). Mean Extraversion score was 30.0 (SD= 5.6). Mean Openness score was 33.8 (SD= 5.0). Mean Agreeableness score was 33.8 (SD= 5.4). Mean Conscientiousness score was 27.0 (SD = 7.3). There was no correlation between any of these personality measures and PIS scores: Neuroticism, [r.sub.s]= .020; Extraversion, [r.sub.s] = .019; Openness, [r.sub.s] = -.050; Agreeableness, [r.sub.s] -.026; Conscientiousness, [r.sub.s] = -.055. To assess whether there was any interaction between personality and expectancy condition with regard to responses to the session questionnaire, for each personality factor, we split participants on the median into high and low groups. We carried out 2 (expectancy condition) x 2 (high or low personality factor) ANOVAs with responses on each of the five session questionnaire items listed in Table 4 as dependent variable. Only one interaction was found to approach significance in these analyses. This was for the Openness personality factor and the qu estion of perceived helping, Openness x Condition, F(1, 56) = 6.51, p= .013. Table 5 gives the mean questionnaire scores that led to the significant interaction. The figures show that low openness participants report higher perceived helping in the positive expectancy compared with the negative expectancy condition, and this trend is reversed for the high openness participants. Given the number of analyses, this interaction can only be regarded as suggestive, particularly because the matching analysis for Study 1 showed no similar trend, F(1, 56) = .379, p = .541.
Self-Report on Influence of Expectancy
Study 2 asked the additional question whether participants believed their confidence of success at the psi task was affected by their knowledge of CW's previous track record. Table 6 gives the percentage of participants who responded "yes," "no," or "unsure" to this question. As the table shows, a small majority of those participants who believed CW had a track record of positive results claimed their confidence was affected (6 of these were believers, 6 were disbelievers). In contrast, a majority of those in the negative expectancy condition claimed their confidence was not affected (10 of these were believers, 10 were disbelievers). The distribution of responses shown in Table 6 is significantly different from chance expectation, according to a chi-square test, [chi square] (2, N=60) = 11.38, p = .003. Therefore, while there seems to be little difference between believers and disbelievers in their responses to this question, there does seem to be an overall difference, suggesting an asymmetry in responses for the one condition compared with the other. This hints at the complexity of expectancy effects, for instance, the possibility that positive and negative expectancy may not lie at different ends of a single continuum but may vary orthogonally to one another.
Self-Report on Nature of Distractions
The session questionnaire in Study 2 included a question asking participants to indicate the nature of their registered distractions. Participants could check more than one category. The majority of distractions were "internal"--either fleeting (such as a brief irrelevant thought, 75%) or longer (such as following an irrelevant train of thought, 78%). 33% of participants reported "external" distractions, and 18% checked the "other" category.
Remote Facilitation of Attention
Study 2, with an effect size of -.11, did not show the predicted remote facilitation of focusing of attention, apparently failing to replicate the previous successful studies by Braud et al. (1995; ES = .25) and Brady and Morris (1997; ES = .27). There are a number of factors that may have caused this failure to replicate. First, there is no psi. If so, this would suggest that the previous studies' results did not represent a genuine psi effect. It is difficult to evaluate this question retrospectively. Instead, we have to continue with future replication attempts under carefully controlled conditions to discover whether remote facilitation of attention continues to occur. Second, perhaps our study did not have adequate statistical power to detect the psi effect, though our power compares favourably with the two previous studies. If we make the generous estimate that the two successful remote focusing studies had an effect size around .3, then Brady and Morris's study with 40 participants had only a power of about .48 to detect this effect at the .05 level, two-tailed (Cohen, 1988). Braud et al.'s study with 60 participants had only a power of about .65 to detect the effect at the .05 level, two-tailed (Cohen, 1988). The present study, which made a one-tailed prediction with 60 participants, had a power of about .76 to detect a remote focusing effect (ES) of .3 at the .05 level on a one-tailed test (Cohen, 1988). We therefore had greater power than the previous two studies, but nevertheless it is still possible that we missed detecting the psi effect because of inadequate power. Third, it is possible that although we made efforts to replicate the procedure of the two previous studies, procedural changes may have been responsible for the failure to replicate. We can only speculate on which of these changes might have been important ones. The present study followed Wiseman and Schlitz's (1997, 1999) procedure of using the experimenter as the remote influencer, as we felt that this might be important for a study of experimenter effects. A similar procedure was used by Braud et al., but Brady and Morris asked participants to bring in friends to act as helper, which may have "boosted" the psi effect in their study. In addition, perhaps the negative expectancy condition in the present study was less "psi-conducive" than in the previous two studies. However, in Study 2, participants in the negative expectancy condition had higher confidence of success and reported more perceived helping than those in the positive expectancy condition. We could speculate further, but systematic research is needed to advance our understanding of what factors are important for success in remote mental influence tasks.
With regard to the effectiveness of the priming article, participants did not know GW in advance of the study; therefore, we feel the article was likely to have had the intended effect. However, a weakness of the present study was that GB, who scheduled the participants and allocated them to conditions, knew some of the participants from previous studies. CB did not do the allocation randomly but attempted to balance the gender between conditions and attempted to alternate allocation to give an even number of positive and negative sessions as the study progressed. Therefore, it is possible (but we feel unlikely) that CB introduced some systematic bias in her allocation to conditions. However, a worthwhile methodological improvement for future studies would be to allocate participants to conditions by a properly random method to eliminate the possibility of such a bias. This methodological improvement was implemented in the follow-up study by Watt and Baker (in press).
Expectancy, Belief, and PIS
We found no indication of higher PIS scores in the positive expectancy condition compared with the negative expectancy condition, and we found no interaction between expectancy, belief, and PIS. Therefore, although there were signs of expectancy effects in some of the questionnaire measures, this did not seem to lead to diverging scores in the psi task, unlike Wiseman and Schlitz's (1997, 1999) studies. This could be because of inadequate statistical power; however, our statistical power is comparable with that of Wiseman and Schlitz's studies. Alternatively, we may have minimized in our study some of the experimenter differences that perhaps contributed to the divergent scoring in the Wiseman and Schlitz studies. For instance, with a single experimenter in the present study, perhaps the influence of experimenter psi is controlled for (as much as it is possible to do so). Also, the experimenter physical and personality characteristics are controlled for in the present study. The main thing that we have attemp ted to manipulate is the participant's belief about the experimenter's track record. Having successfully isolated this variable, our results suggest that this is only a small contributory factor to parapsychology's experimenter effect.
Belief and Expectancy
In both studies, participants' responses to questions about psi belief did not seem to be affected by the priming article. The participants completed their belief questionnaires prior to any contact with the experimenter. It is likely that completing a belief questionnaire following the presession chat with an experimenter may be even more likely to influence belief responses, as was found by Wiseman and Schlitz (1997, 1999). This is because a one-to-one "pep-talk" gives more opportunity for the experimenter to communicate, verbally and nonverbally, their expectations for success and their opinions about psi. Also, it is possible that in our studies, even the negative expectancy priming article emphasized the importance of DMILS research, which might be a more positive slant than a strongly skeptical article or experimenter might take.
Confidence, Perceived Helping, Experimenter Evaluations, and Expectancy
There were no significant differences between expectancy conditions in how participants responded to items in the session questionnaire. However, the strongest trend in Study 1 (that participants in the positive expectancy condition gave CW higher ratings for her ability to instill confidence in the task compared with those in the negative expectancy condition, z = 1.213) was found to a slightly stronger degree in Study 2 (z = 1.685). Post hoc, combining the results for this question in Study 1 and Study 2 for greater statistical power, we find Mann--Whitney z = 2.038, p = .042, two-tailed. The combined results for each of the other four questions (prior confidence, perceived helping, experimenter warmth, and experimenter professionalism) gave only small zscores (.366, .561, .731, and .259, respectively). Therefore, the experimenter's ability to instill confidence seems to be the only question on which positive and negative expectancy participants could be differentiated, suggesting it may be relatively more important in experimenter effects than the other questions.
Experimenter Effects and Artifact Detection
As this study is about experimenter effects, it is interesting to note that the reason that we made substantial efforts to identify whether the psi results in Study i were artifactual was because we felt the apparent psi results were unusually strong, and were therefore unexpected. If the remote helping effect size had been of a similar magnitude to the previous two remote helping studies, we might have reacted quite differently. In that case, we might have regarded our results as successfully replicating the effect size of the previous two studies, and we might not have made any further effort to check whether the results were valid. This would have led to dissemination of a spurious positive result. We have no way of evaluating whether the previous remote helping studies also carry undetected artifacts, and we have no wish to suggest that this might be the case. However, our present experience is at the very least a cautionary tale. We should also note that the experimenter's expectancy cuts both ways, so t hat a skeptical experimenter, who gets null results as expected, may similarly assume that his or her study results are valid, when in fact a procedural flaw may be operating to obscure a positive psi effect. Our experience is a graphic example of how one kind of experimenter expectancy effect may operate to produce results in line with the experimenter's expectations.
TABLE 1 MEAN (AND STANDARD DEVIATION) SESSION QUESTIONNAIRE RESPONSES, BY CONDITIOIN Question Positive Negative z p Expectancy Expectancy Confidence of success 3.1 (0.71) 2.9 (0.86) 1.066 .29 Perceived helping 3.3 (0.75) 3.0 (1.1) 1.150 .25 Experimenter warmth 4.8 (0.41) 4.7 (0.64) 0.064 .95 Experimenter professionalism 4.7 (0.54) 4.6 (0.67) 0.282 .78 Experimenter instilling confidence 4.4 (0.50) 4.2 (0.79) 1.213 .23 Note. The z scores are based on Mann-Whitney comparison. p values are two-tailed. TABLE 2 SPEARMAN CORRELATION COEFFICIENTS BETWEEN PERSONALITY FACTORS AND SESSION QUESTIONNAIRE RESPONSES Question N E O A C Confidence of success .236 -.119 .043 -.016 -.022 Perceived helping -.065 .171 -.230 .092 .096 Experimenter warmth -.216 .208 -.303 .139 .113 Experimenter professionalism .080 .096 -.286 .082 .086 Experimenter instilling confidence -.062 .337 -.170 .101 .242 Note. N = Neuroticism E = Extraversion O = Openness A = Agreeableness C = Conscientiousness. TABLE 3 MEAN (AND STANDARD DEVIATION) OF REGISTERED DISTRACTIONS (BUTTON PRESSES) AND PIS SCORES, BY BELIEF AND CONDITION Positive expectancy Negative expectancy Group Help Control PIS Help Believers 11.54 11.46 .498 12.65 (3.9) (4.4) (0.1) (8.3) Disbelievers 12.94 13.18 .512 13.08 (9.4) (9.0) (0.1) (10.5) All participants 12.33 12.43 .506 12.83 (7.7) (7.6) (0.1) (9.5) Negative expectancy Overall Group Control PIS Help Control PIS Believers 11.53 .488 12.17 11.5 .492 (6.4) (0.1) (6.8) (5.6) (0.1) Disbelievers 12.54 .492 13.0 12.9 .504 (10.3) (0.1) (9.9) (9.6) (0.1) All participants 11.97 .490 12.58 12.20 .498 (8.5) (0.1) (8.6) (8.0) (0.1) Note: PIS = percentage influence score. TABLE 4 MEAN (AND STANDARD DEVIATION) SESSION QUESTIONNAIRE RESPONSES, BY CONDITION Question Positive Negative z p Expectancy Expectancy Confidence of success 2.8 (0.97) 2.9 (0.83) -0.564 .57 Perceived helping 2.9 (1.16) 3.1 (0.87) -0.436 .66 Experimenter warmth 4.9 (0.35) 4.8 (0.43) 0.993 .32 Experimenter professionalism 4.8 (0.43) 4.7 (0.52) 0.070 .94 Experimenter instilling 4.5 (0.63) 4.2 (0.71) 1.685 .09 confidence Note: The z scores are given based on Mann-Whitney comparison of positive and negative expectancy conditions; p values are two-tailed. TABLE 5 MEAN (AND STANDARD DEVIATION) RESPONSES ON QUESTION OF PERCEIVED HELPING, ARRANGED BY EXPECTANCY CONDITION AND OPENNESS SCORING Condition Low Openness High Openness Positive expectancy 3.4 (0.72) n = 16 2.4 (1.33) n = 14 Negative expectancy 2.9 (1.07) n = 14 3.2 (0.65) n = 16 TABLE 6 PERCENTAGE OF PARTICIPATNS RESPONDING "YES," "NO," OR "UNSURE" TO QUESTION ON INFLUENCE OF EXPECTANCY Overall Positive Negative expectancy expectancy Yes 23.3 20.0 3.3 No 48.3 15.0 33.3 Unsure 28.3 15.0 13.3
(1.) A referee of this article helpfully pointed out that the use of PIS scores had some limitations because the PIS score is not standardized by the standard deviation (see Schmidt, Schneider, Binder, Burkle, & Walach, 2001). While acknowledging this limitation, we did not change our planned analyses because we had concerns about the reliability of a more complex session outcome measure such as a paired t test for the eight pairs of epochs in each session. This measure might be unreliable because of the small number of epochs and because many of the cells would have only zero, one, or two button presses. Also, a second t test would have to be performed on the per-session measure, which would tend to lower the power of the analysis. The outcome measure for Hypothesis 1--a t test with 60 pairs of data points (each participant's total number of help presses and total number of control presses)--was chosen because it was felt this test would be both reliable and appropriate.
BRADY, C., & MORRIS, R. (1997). Attention focusing facilitated through remote mental interaction: A replication and exploration of parameters. Proceedings of Presented Papers: Parapsychological Association 40th Annual Convention, pp. 73-91.
BRAUD, W. G., & SCHLITZ, M.J. (1991). Conscious interactions with remote biological systems: Anomalous intentionality effects. Subtle Energies, 2, 1-46.
BRAUD, W., SHAFER, D., McNEILL, K., & GUERRA, V. (1995). Attention focusing facilitated through remote mental Interaction. Journal of the American Society for Psychical Research, 89, 103-115.
COHEN, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: Erlbaum.
COSTA, P. T., & McCRAE, R. R. (1991). Neo Five-Factor Inventory. Odessa, FL: Psychological Assessment Resources.
HONORTON, C., RAMSEY, M., & CABIBBO, C. (1975). Experimenter effects in extrasensory perception. Journal of the American Society for Psychical Research, 69, 135-139.
KENNEDY, J. E., & TADDONIO, J. L. (1976). Experimenter effects in parapsychological research. Journal of Parapsychology, 40, 1-33.
LAYTON, B. D., & TURNBULL, B. (1975). Belief, evaluation, and performance on an ESP task. Journal of Experimental Social Psychology, 11, 166-179.
PALMER, J. (1997). The challenge of experimenter psi. European Journal of Parapsychology, 13,110-125.
PARKER, A., FREDERIKSEN, A., & JOHANSSON, H. (1997). Towards specifying the recipe for success with the ganzfeld. European Journal of Parapsychology, 13, 15-27.
ROSENTHAL, R., & ROSNOW, R. L. (1991). Essentials of behavioral research: Methods and data analysis. New York: McGraw-Hill.
SCHMIDT, S., SCHNEIDER, R., BINDER, M., BURKLE, D., & WALACH, H. (2001). Investigating methodological issues in EDA-DMILS: Results from a pilot study. Journal of Parapsychology, 65, 59-82.
SCHNEIDER, R., BINDER, M., & WALACH, H. (2000). Examining the role of neutral versus personal experimenter-participant interactions: An EDA--DMILS experiment. Journal of Parapsychology, 64, 181-194.
TADDONIO, J. L. (1975). Attitudes and expectancies in ESP scoring. Journal of Parapsychology, 39, 289-296.
WATT, C., & BARER, I. S. (in press). Remote facilitation of attention focusing with psi-supportive versus psi-unsupportive experimenter suggestions. Journal of Parapsychology.
WISEMAN, R., & SCHLITZ, M. (1997). Experimenter effects and the remote detection of staring. Journal of Parapsychology, 61, 197-207.
WISEMAN, R., & SCHLITZ, M. (1999). Experimenter effects and the remote detection of staring: An attempted replication. Proceedings of Presented Papers: Parapsychological Association 42nd Annual Convention, pp. 471-479.
We are grateful to the Perrott-Warrick Fund and the Koestler Parapsychology Unit for providing financial support for these studies. We would like to thank Dr Paul Stevens for his computing support, Prof Bob Morris and Dr Richard Wiseman for their helpful cornments on an earlier version of this article, and Dr Richard Wiseman for helping to conduct some procedural checks for study 1. An abbreviated version of this paper was presented at the 2001 Annual International Convention of the Parapsychological Association. We are grateful to our anonymous referees for their helpful comments.
Department of Psychology
University of Edinburgh
7 George Square
Edinburgh EH8 9JZ
|Printer friendly Cite/link Email Feedback|
|Author:||Watt, Caroline; Brady, Claire|
|Publication:||The Journal of Parapsychology|
|Date:||Mar 1, 2002|
|Previous Article:||Parapsychology and transpersonal psychology: "Anomalies" to be explained away or spirit to manifest?|
|Next Article:||The ganzfeld debate continued: A response to Milton and Wiseman (2001).|