JOURNAL OF EXPERIMENTAL SOCIAL PSYCHOLOGY 11, 166-179 (1975)
Belief, Evaluation, and Performance on an ESP TaskL2 BRUCE D . LAYTON:' AND BILL T U R N B U L L
State University o f New York at Albany and University o f British Colnmbia Received March 7, 1972 Belief in the possibility of demonstrating ESP on an experimental task and evaluation of ESP were independently manipulated in two laboratory studies investigating the relationship between these variables and ESP task performance. The results of Experiment 1 indicated that subjects exposed to a positive evaluation of ESP scored higher on the ESP task than those exposed to a negative evaluation of ESP. Furthermore, a significant belief x sex of subject interaction was found. The interaction indicated that males exposed to the disbelief manipulation scored significantly higher on the task than males exposed to the belief manipulation or females exposed to the disbelief manipulation. In general, the ESP task performance results for females followed the pattern predicted from the Ajzen and Fishbein (1972) attitude-behavior model, while those for males did not. Experiment I1 was a strict methodological replication of Experiment 1. In Experiment II there were no significant effects present upon ESP task performance. The implications of these contradictory results for ESP research specifically and significance levels generally are both discussed.
In any field of scientific investigation there is often considerable controversy over the demonstration of some "new" phenomenon. Those who would not be convinced of its existence, regardless of the evidence presented, remain unconverted, while those convinced become adamant. Seeking acceptance in the scientific community, proponents of the "new" phenomenon concentrate their energies upon demonstrating the actuality of the phenomenon rather than advancing by utilizing explanatory constructs and testable theories. We believe that this is particularly true of research in parapsychology. In order to show that psi (an incluEditor's note. This article was initially submitted as a manuscript reporting the first of the two experiments described here. Because of the unusual nature of the research and equivocal nature of the results, the authors and editor entered into an agreement: the authors agreed to conduct an exact replication of the initial experiment and to report the findings, regardless of outcome; the editor agreed to publish a paper reporting both sets of results, regardless of the outcome of the second study. '-' We wish to thank John Thibaut, Nehemia Friedland, and James Carpenter for their helpful comments and suggestions on a preliminary version of the manuscript, and Bob Frey for his help with the statistical analyses. :~ Requests for reprints should be sent to Bruce D. Layton, Psychology Department, State University of New York, 1400 Washington Avenue, Albany, New York 12222. 166 Copyright9 1975 by AcademicPress, Inc. Printedin the UnitedStates. All rightsof reproductionin any formreserved.
BELIEF, EVALUATION, AND ESP
167
sire term for all parapsychological phenomena) exists, parapsychological research has become mired in a seemingly relentless search for astronomical levels of significance. It is our contention that there is enough evidence to warrant considerable confidence in the existence of psi (cf. recent volumes of the Journal of Parapsychology and the Journal of the American Society for Psychical Research). Research which only documents the reality of psi phenomena, however, is at best merely descriptive. What is needed are theories which can incorporate the existing empirical data and generate specific testable predictions. We believe that the experimental methodology of psychological research should be applicable to parapsychological phenomena. Thus we are claiming that the laws and variables of psi are qualitatively and quantitatively similar to those of "normal" psychological phenomena. As social psychologists, our tendency is to manipulate a variable that will affect an individual's perception or understanding of a given stimulus. In the present experiment, we will focus upon the manipulation of attitude. There has been a lengthy and intensive debate in the social psychological literature regarding the relationship between actual behavior and attitude toward an object (LaPiere, 1934; Campbell, 1961; McGuire, 1969; Ajzen, 1971). This relationship has also been of interest to many parapsychologists. Although a review of the social psychological literature reveals a minimal correlation between attitude and behavior, a similar review of the parapsychological literature indicates a much stronger relationship. Studies by Bhadra (1966), Bevan (1947), Casper (1951), Eilbert and Schmeidler (1950), Schmeidler (1943), Van de Castle and White (1955), and Woodruff and Dale (1950) have been interpreted as indicating a relationship between "attitude" toward psi and performance on a parapsychological task. Generally, those subjects who believed in ESP (sheep) scored significantly higher on ESP tasks than those who did not believe in ESP (goats) (cf. (Palmer, 1971, 1972) for a review). Given the extensive literature on the relationship of "attitude" and psi, it may seem odd that we should attempt another experiment. There are, however, two distinct reasons for this study. Because previous studies have been correlational in nature, the existing literature indicates only the presence of a relationship between "attitude" and psi, leaving the causal direction unspecified. In order to determine whether attitudes influence behavior, we must manipulate attitudes and measure the effects on behavior. The second purpose of this experiment is to clarify the concept of "attitude." In most parapsychological research, the sheep-goat distinction has measured a belief in the existence of ESP (e.g. (Musso, 1965)). Other studies, however, have defined the dimension as encompassing a positive or negative evaluation of ESP (e.g. (Ryzl, 1968)). Still other
168
LAYTON AND TURNBULL
studies have combined a belief in the existence of ESP with an evaluation of ESP into a unitary sheep-goat index (e.g. (Stanford, 1965)). Fishbein and Raven (1962) have demonstrated, however, that a belief in the existence of X can be orthogonal both theoretically and empirically to an evaluation of X. Any examination of the relationship between these variables and psi must conceptually and operationally distinguish between belief and evaluation. The present experiment employs a 2 x 2 x 2 factorial design which incorporates the distinction between belief in the existence of ESP and evaluation of ESP by including these variables as independent factors. The belief manipulation is accomplished by an experimenter who professes either to believe or disbelieve in the existence of ESP. The evaluation manipulation is created by the experimenter's purported positive or negative evaluation of ESP. The third factor is sex of subjects. A control group in which no manipulation of belief or evaluation occurs was included as a replication of previous work. Performance on a clairvoyance task served as the primary dependent variable.
Predictions Our predictions are based on the Ajzen and Fishbein (1969, 1970, 1972, 1973; Ajzen, 1971) model of behavioral intentions. In this model, a behavioral intention (BI) is a weighted, joint function of the attitude toward performing a given behavior in a given situation (A-act) and the product of social normative beliefs (NBs) and the motivation to comply with these normative beliefs (Mcs). A-act, in turn, is a product of the beliefs about the consequences of performing the act and the evaluation of these consequences. Behavioral intentions are assumed to mediate overt behavior. In the present experimental situation, the value of A-act is manipulated by capitalizing upon the experimenter's credibility and ostensible expertise in both psychological and parapsychological research. Thus, the value of A-act is dependent upon the experimenter's claim that ESP would (or would not) be demonstrated in the situation, and upon his evaluation of ESP. Both the social normative beliefs and the motivation to comply with these beliefs are manipulated as byproducts of the direct manipulation of A-act. Because the experimenter is the relevant reference person for the situation, normative behavior for a subject in this situation is to agree with the experimenter's views of ESP (cf. (Orne, 1962)). When the Ajzen and Fishbein model is applied to our specific conditions, we are able to make some quantitative predictions. If large values of A-act represent a belief in the existence of ESP and a favorable evaluation of it, then the magnitude of A-act should be largest for subjects in
BELIEF, EVALUATION, AND ESP
169
the Sheep-Positive condition. Similarly, it should be smallest in the Goat-Negative condition, and intermediate in the other two conditions. Predicting the value of the social normative belief component, however, is more complex. Subjects in the Sheep condition are led to believe that the experimenter expects ESP to occur, while those in the Goat condition are led to believe that it will not. This is complicated, however, by the manipulated evaluation of ESP. Subjects exposed to a positive evaluation of ESP should feel it is good if they can demonstrate ESP. Subjects presented with a negative evaluation should feel that it is bad if they demonstrate ESP. Thus the various evaluation and belief combinations present consistent social normative beliefs (what should be done, what it is good to do) only in the Sheep-Positive and Goat-Negative conditions. Subjects in the other two conditions are presented with ambiguous, conflicting normative information. If we assume that subjects in all conditions are equally motivated to comply with the social normative beliefs of the situation, then the product of social normative beliefs and motivation to comply with these beliefs should be highest for SheepPositive subjects, lowest for Goat-Negative subjects, and intermediate for Sheep-Negative and Goat-Positive subjects. Thus, from the Ajzen and Fishbein model we would predict that subjects in the Sheep-Positive condition should score significantly higher on the ESP task than subjects in the Goat-Negative condition. Subjects in the Sheep-Negative and Goat-Positive conditions should score at an intermediate position. Because females are generally more responsive to demand characteristics than males (cf. (Silverman, 1968)), we predict that this basic pattern of scores will be stronger for females than males. Although not specifically mentioned above, main effects of belief and evaluation on ESP scoring are also predicted from an application of the Ajzen and Fishbein model. METHOD
Subjects The subjects were unselected volunteers at the University of North Carolina who signed up for a "matching experiment" as partial fulfillment of the experimental requirements for the introductory psychology course. One hundred males and 79 females participated in the experiment. Subjects were run in groups of 5-15.
Independent Variables The design included three independent variables: experimenter's purported belief in the existence of ESP both on the matching task and in general, experimenter's purported evaluation of ESP, and sex of subjects. The experimenter portrayed himself as either believing or not believing in the existence of ESP and as holding a positive or negative evaluation of ESP. Both the belief and evaluation manipulations were present in the instructions given to the subjects prior to the administration of the ESP task.
170
LAYTON AND TURNBULL
Procedure When all subjects for a given session had entered the room, the experimenter introduced himself as a postdoctoral student at the University of North Carolina who was involved with a research group studying extrasensory perception. Both his status as a postdoctoral student with experience in ESP research and his well-dressed appearance were intended to create an impression of expertise and credibility in the eyes of the subjects. Although a description of clairvoyance research was given, the experimenter noted clearly that the ESP task in this experiment was representative of ESP in general. He then gave a detailed description of the ESP task in which each subject was to match a separate list of 100 randomly generated digits using the digits "1," "2," "3," "4," and "5." The experimenter stressed the difficulties inherent in reports of naturalistic instances of ESP and explained the necessity for conducting carefully controlled laboratory studies in order to decide whether or not ESP exists. The research project which he was involved in, he added, had been specifically designed so that the results could be unambiguously interpreted. To date, nineteen studies in this project had supposedly been completed. At this point the belief manipulation was introduced. Subjects in the Sheep conditions were told that previous research had demonstrated the existence of ESP. Furthermore, they were informed that the experimenter was personally convinced both that ESP existed generally and that it would be demonstrated in the matching task. Subjects in the Goat conditions heard opposite instructions that stressed the nonexistence of ESP. The manipulation of evaluation of ESP occmTed immediately after the belief manipulation. In general, it consisted of the experimenter's suggesting either possible beneficial or harmful effects of the use of ESP. The examples included were collected fl'om a survey of pretest subjects and psychology graduate students on the "'good and bad things about ESP." A control group was included for comparison with the existent correlational literature. The procedure for the controls was identical to that of the experimental groups except that the experimenter neither mentioned nor attempted to manipulate belief in or evaluation of ESP. Following the manipulations, subjects were administered "Questionnaire Series I," a booklet containing various scales for rating ESP, the experimenter, and the self. These forms are described more fully below, Upon subjects' completion of the questionnaire, the experimenter once again explained the procedure to be followed on the matching task until all subjects appeared to understand what was required of them. Subjects then proceeded with the task. They were under no time constraints. Responses on all experimental forms were anonymous; no names were ever requested by the experimenter. When all subjects had finished the matching task, they were asked to complete the final questionnaire which assessed their confidence in both their own and other subjects' performance on the ESP matching task, At this point the experiment was terminated, and subjects were completely debriefed as to the exact nature and purpose of the experiment. The necessity for including deceptions was explained, and all questions were answered. All subjects agreed not to reveal the nature of the experiment.
Dependent Variables The first dependent assessments were completed immediately following the induction of the experimental manipulations and prior to the administration of the ESP task. In addition to giving some biographical information in a booklet entitled "Questionnaire Series 1,'" subjects rated each of six concepts related to ESP on five 15-point semantic differential scales. Four of these concepts assessed a subject's belief in the existence of ESP. These included: "Some people have ESP ability": "1 (referring to yourself) have ESP ability":
BELIEF, EVALUATION, AND ESP
171
"'In the present experimental situation of a matching task, some people will demonstrate ESP ability"; "In the present experimental situation of a matching task, 1 (referring to yourself) will demonstrate ESP ability." Each of these concepts was rated along five dimensions of the Fishbein and Raven (1962) belief scale. The two remaining concepts referred to the subjects' evaluation of or attitude toward ESP. These included: "Give your evaluation of the concept ESP" and "How do you personally feel about people who have ESP ability." Each of these latter two concepts was rated along the Fishbein and Raven (1962) attitude dimensions. In "Questionnaire Series 1" subjects were also asked to indicate their feelings about the experimenter on 27 7-point semantic differential scales. These included dimensions of the experimenter's believability, competence, likability, etc. The final assessments in the booklet were a set of 30 7-point semantic differential scales on which the subjects were asked to rate their present feelings on items such as tense-relaxed, creative-uncreative, sincere-insincere, etc. The principal dependent measure was the "matching task" itself. Subjects were given an opaque envelope (11 x 12 in. ) with a blank response sheet stapled to the front. The response sheet had four columns of 25 blanks numbered consecutively from 1-100. On a similar sheet inside each envelope was the target list (and on another sheet its carbon copy) for that particular subject. The target list was composed of 100 individual digits which were arranged in a random order generated by an IBM 360 computer. One-hundredseventy-nine random samples of 100 one digit numbers using the digits from one through five inclusive were drawn with replacement. One of the authors compiled the target samples and together with a research assistant transcribed the target orders onto the response sheets. The other author instructed the subjects in the clairvoyance task to indicate what they believed each of the 100 digits to be by writing the appropriate digit in the corresponding space provided on the envelope's cover sheet. Although subjects were not informed of their success on the clairvoyance task, upon completion of the matching task they were given a three-item questionnaire assessing feelings about their performance. Subjects responded on 7-point scales to questions on how well they felt they did on the matching task, how well they felt other people did, and how confident they felt about their answers on the matching task.
RESU LTS
Manipulation Checks Subjects' responses to the five semantic differential belief scales were scored such that higher numbers indicated a greater belief in the existence of ESP. When these scores were summed separately for each of the four belief concepts, a least squares, multivariate analysis of variance indicated a successful induction of the belief manipulation (Multivariate F ( 4 , 1 6 3 ) = 20.31, p < .001). Because A-act from the Ajzen and Fishbein model reflects situational aspects of belief, a more meaningful manipulation check would involve only the question referring to a belief in one's own demonstration of ESP in the matching task situation. The univariate F for the belief manipulation was significant upon the summed scales for this item ( F ( 1 , 1 6 6 ) = 29.16, p < .001). Subjects in the Sheep conditions ()~ = 43.24) believed it was more likely that they would demonstrate E S P on the matching task than did subjects in the G o a t conditions 0~ = 29.72).
172
LAYTON AND TURNBULL
A least squares, multivariate analysis of variance also indicated a successful induction of the evaluation manipulation upon the summed scales within each attitude item (Multivariate F ( 2 , 1 6 5 ) = 27.32, p < .001). The most relevant attitude item assessed subjects' evaluation of ESP. The significant main effect upon this variable ( F ( 1 , 1 6 6 ) = 5 0 . 0 6 , p < .001)indicated that subjects in the positive evaluation conditions (X = 60.11 ) were more favorably disposed toward ESP than subjects in the negative conditions (X = 47.00). Although the evaluation manipulation did not significantly affect any of the belief items, there was a surprising main effect for the belief manipulation upon subjects' rated evaluation of ESP ( F ( 1 , 1 6 6 ) = 7.06, p < .009). Subjects in the Sheep conditions (~" = 55.98) rated ESP more favorably than subjects in the Goat conditions (.~ = 55.13). Contrary to Fishbein and Raven's (1962) assertion, this latter effect suggests that the Attitude (evaluation)-Belief Scales may not be tapping completely orthogonal dimensions.
Performance on the ESP Task A subject's answers to the ESP matching task were compared with the original list contained in each envelope. A "hit" was scored if the digit written by the subject in a particular slot was the same as the digit in that slot on the sheets inside the envelope. The matching task was first scored by a research assistant who was blind both to the experimental condition of each subject and to the specific hypotheses of the experiment. Each "hit" was checked off, and the score summed for each of the four columns separately and for the entire task. A score was computed which reflected the subject's signed deviation from chance performance. Because the probability of a correct response occurring by chance is 1/5, the mean chance expectation in each column of 25 trials is 5 correct. Similarly the mean chance expectation for the entire task of 100 trials is a score of 20 correct responses. Each subject's performance was also scored by one of the authors using the carbon copy of the subject's target order which had also been placed inside the envelope. The same scoring procedure as above was followed, and any discrepancies between the research assistant's scoring and those of the authors were noted and corrected. The analyses to be reported here were conducted upon the deviations from chance expectation of the correct scores for the entire task. Because deviation scores are used, the mean chance expectation for this task is zero correct. Effect of belief in the existence of ESP upon task performance. A least squares analysis of variance indicated that the manipulated belief in the existence of ESP did not have any significant main effects upon performance on the matching task (F < 1). Although the belief manipula-
173
B E L I E F , E V A L U A T I O N , A N D ESP
tion did not influence actual performance, it did affect subjects' perceptions o f how well they believed they did (F(1,169) = 27.65, p < .001), how well they believed other people did ( F ( 1 , 1 6 9 ) = 16.78, p < .001), and how confident they were of their answers on the matching task (F(1,169) = 10.57, p < .001). Subjects exposed to the Sheep instuctions were more confident both of their own and other people's ability than were subjects in the G o a t conditions. Although there was no main effect of the belief manipulation upon performance on the matching task, this manipulation significantly interacted with the sex of subject factor (F(1,169) = 4.86, p < .029). The mean deviation scores for this interaction can be seen in Table 1. Descriptively, males in the G o a t conditions and females in the Sheep conditions have ESP scores greater than the mean chance expectation of zero, while males in the Sheep conditions and females in the G o a t conditions have scores less than the mean chance expectation. Thus the pattern of results for females replicates earlier studies in that Sheep score higher than Goats. Males, however, show the opposite pattern. Furthermore, there is a significant sex difference within the G o a t conditions ( F ( 1 , 1 6 9 ) = 5.29, p < .023). Although the negative mean deviation score for females in the G o a t conditions is not significantly different from chance ( F ( 1 , 1 6 9 ) = 1.76, p < .18), the positive mean deviation score for males in the G o a t conditions is significantly different from chance (F(1,169) = 4.46, p < .036). Effect of evaluation of ESP upon task performance. The predicted effect of the evaluation manipulation on ESP performance was marginally significant as a two-tailed test ( F ( 1 , 1 6 9 ) = 3.62, p < .059). Subjects in the positive evaluation condition had a mean deviation from chance o f +.67. Subjects exposed to a negative evaluation of ESP, however, had a
TABLE 1 MEAN DEVIATION FROM CHANCE EXPECTATION ON THE MATCHING TASK Belief Sheep
Goat
Evaluation
Male
Female
Male
Female
Positive Negative Average
--.04 --.85 --.51
+ 1.40 --.78 +.27
+ 1.52 +.75 + 1.17
--.64 -- 1.06 --.79
Note. Higher numbers indicate greater deviation from the mean chance expectation of zero correct.
174
LAYTON AND TURNBULL
mean deviation score o f - . 5 0 . Although the order of these means was in the predicted direction, neither of them was significantly different from the mean chance expectation for the matching task. Evaluation did not interact significantly either with the belief or sex of subject factors. The three-way interaction also failed to reach significance. There were no effects of evaluation of ESP upon the various confidence ratings (F's < i). Controls We had originally hoped to use control subjects who were not exposed to any experimental manipulation as a replication of some of the early correlational studies involving a Sheep-Goat distinction. Relative to the levels of belief and evaluation for our experimental subjects, however, virtually all the controls would be classified as Sheep-Positive. Thus any post hoc division of the controls along the belief and evaluative dimensions would have been purely arbitrary since no natural cutting point presented itself. If the basis for discrimination were the sex of subjects factor, however, the control data would descriptively resemble the Sheep-Positive experimental data in that control males (X = - 1 . 0 4 ) scored less than control females ()~ = +.85). The difference between the means was not significant, however. Discussion The results of the present experiment suggest the utility of a scientific approach to an examination of parapsychological phenomena. Using a design and methodology typical of experimental social psychology, we manipulated subjects' belief in and evaluation of ESP. Neither the experimenter nor (so far as we know) the subjects were psychics, and no attempt was made to create any special mystical atmosphere. In fact, the experiment was run like other social psychological experiments except that rather stringent controls were implemented in order to eliminate opportunities for faking, cheating, or general dishonesty on the part of either the subjects or experimenter. Despite these rigorous methodological restrictions, the results clearly indicated significant differences between conditions on an ESP task. Since belief and evaluation were manipulated, and since the design of the study seems to control for extraneous variables such as population response biases, we can be fairly confident in claiming that subjects' belief and evaluation of ESP influenced their performance on the ESP task. The importance of the distinction between belief and evaluation is attested to by our results. Subjects exposed to a positive evaluation of ESP rated ESP more positively and scored significantly higher on the ESP task than those exposed to a negative evaluation of ESP. The
175
B E L I E F , E V A L U A T I O N , A N D ESP
results obtained for the belief manipulation, however, were not as predicted. While subjects led to believe that E S P was possible in the experimental situation (Sheep) did evidence greater belief than those led to believe that E S P was impossible in the situation (Goats), no differences were obtained on the ESP task itself for this manipulation. The most surprising result of the present experiment is the significant belief by sex interaction on ESP scores. Broken down, this interaction indicates that male Goats scored significantly higher than both male Sheep and female Goats. In general, the pattern of results for females is that which we predicted from the Ajzen and Fishbein model, SheepPositives scoring higher than Goat-Negatives with the other two conditions intermediate. The results for males do not follow the predicted pattern for belief but do so for evaluation. H o w are we to account for the results of our male subjects? One possibility is that the belief manipulation did not have the desired effect on males. This would mean, of course, that the belief component of A-act was differentially manipulated across sex. H o w e v e r , as can be seen in Table 2, the belief means for both males and females fall in the predicted pattern. Thus differences in the belief component of A-act do not appear to mediate the sex interaction. Furthermore, if the interaction were mediated by differential change in either the social normative beliefs or the motivation to comply, then a significant evaluation by sex interaction on the E S P scores should also be found. This interaction, however, was not significant. In short, there appears to be no ready explanation of the belief by sex interaction. T h e r e are, however, two remaining possibilities. One is that males and females respond differently to the specific type of task used and that this differential responding does not evidence itself as a main effect but rather as an interaction. T h e second, and more troublesome possibility, is that the significant results may represent a T y p e I error, that is, they may be due to chance. Because of this latter possibil-
TABLE 2 MEAN BELIEF SCORES FOR ALL EXPERIMENTAL SUBJECTS Female
Sheep Goat
Male
Positive
Negative
Positive
Negative
44.67 28.64
42.87 29.71
41.74 34.74
45.15 23.19
Note. Higher scores indicate an individual's greater belief that he will d e m o n s t r a t e ESP in the matching task situation. Scores are s u m s of 5 semantic differential scales. T h e s u m s have a possible range of 5-75.
176
LAYTON AND TURNBULL
ity of rejecting a null hypothesis that was in fact true, a second experiment was conducted. The second experiment was an exact methodological replication of the first experiment. EXPERIMENT
II
Subjects and Procedure Unselected, volunteer subjects were drawn from the same type of subject pool at the University of North Carolina as in Experiment 1. To maximize comparability with the original sample, Experiment Ii was run at the same time in the school year as Experiment I but a year later. Two-hundred-thirty-five subjects were run in the replication. Of these, 105 were males, and 130 were females. The procedure was an exact replication of that from Experiment I. The same experimenter presented the same cover stories containing the belief and evaluation manipulations. Subjects completed the manipulation checks, self-ratings, and ratings of the experimenter before proceeding to the matching task. For the matching task, 235 new random samples of 100 digits were generated by the computer using a different entry point from that of Experiment [.
RESULTS
The data were analyzed as in Experiment 1. The manipulation checks revealed a satisfactory induction of both the belief and evaluation manipulations. The univariate F for the belief manipulation upon the summed scales of the item referring to a belief in one's own demonstration of ESP in the matching task situation was significant (F(1,225)= 32.27, p < .001). Subjects in the Sheep conditions (X = 39.48) believed it was more likely that they would demonstrate ESP on the matching task than did subjects in the Goat conditions (X = 27.49). As in Experiment I, the belief manipulation was significant upon all the belief manipulation checks individually and collectively (Multivariate F(4,222)= 21.52, p < .001). The multivariate F for the evaluation manipulation upon the attitude items was also significant (Multivariate F ( 2 , 2 2 4 ) = 41.78, p < .001). The significant univariate F for this manipulation upon subjects' evaluation of ESP (F(1,225) = 61.26, p < .001) revealed that subjects in the positive evaluation conditions (X = 61.50) rated ESP more favorably than those in the negative conditions (X = 49.86). Although the evaluation manipulation did not significantly affect any of the belief items, there again was a main effect for the belief manipulation upon subjects' rated evaluation of ESP (F(1,225)= 19.58, p < .001). Unlike Experiment I, however, the belief manipulation also affected subjects' evaluations of people who had ESP ability (F(1,225) = 15.02, p < .001). For both variables in Experiment II, subjects in the Sheep conditions were more favorably disposed toward ESP than were subjects in the Goat conditions.
BELIEF, EVALUATION, AND ESP
177
T h e effects of the manipulated variables upon performance on the matching task can be summarized rather succinctly: no significant effects were present. The results from Experiment I failed to replicate in Experiment II, at least for subjects in the experimental conditions (all F's < I). For the control subjects the means were in the same direction as Experiment 1 with males ( J ( - - - - . 3 5 ) scoring lower than females (J( = + 2.31 ). This difference was marginally significant (F(1,225) = 2.92, p < .1).
GENERAL DISCUSSION A major and persistent criticism of psi research has been the lack of repeatability of results. The present investigation also failed the test of repeatability. A strict methodological replication of the first experiment yielded no significant effects of the experimental manipulations upon ESP performance. Given the contradictory results, a possible interpretation is that the true state of the world is one in which psi, as defined in these studies, does not exist. The apparent occurrence of psi effects in the first experiment is thus interpreted as a T y p e 1 error. It seems, bowever, that some principled reason should be given for rejecting only the results of the first experiment but not those of the second. Many other similar studies have obtained results indicative of psi; furthermore, failures to replicate psychological effects other than psi typically have not led to the rejection of the results on the grounds that they represented T y p e I errors. F o r example, in a long series of experiments designed to uncover the necessary and sufficient conditions of cognitive dissonance, Collins and his associates (cf. (Collins, Ashmore, Hornbeck, and Whitney, 1970)) failed to find a paradigm that reliably produced dissonance effects. The literature of dissonance effects is not consequently regarded, however, as a literature of T y p e I errors. T h e r e appear to be two general reasons for rejecting the results of the first experiment. Due to the a priori unlikelihood of psi effects, the occurence of a statistically significant psi effect needs to be weighed against the unlikelihood of that effect. Thus, it may be argued that the apparent occurrence of any psi effect has a high probability of representing a T y p e I error. Secondly, if psi effects exist, the serious and far reaching implications for the causal influence model may likewise suggest that great caution be exercised in accepting such results. Taken together, these two reasons lead to the conclusion that stricter criteria of acceptance should be applied to psi than to nonpsi research. Thus, one may demand that psi data be accepted only if replicated and/or that a more stringent criterion of statistical significance than the commonly accepted .05 level be applied to psi research. And on either of these criteria, the present investigation failed to demonstrate the existence of psi.
178
LAYTON AND TURNBULL
A second interpretation of our results is that psi did occur in the first study but not the second because the two experimental situations differed in important ways. For example, the experimenter's knowledge of the results of the first study, the increased fluidity of his presentation, or any changes in the subject population could all have contributed to the contradictory results. Of course, there are always differences between "identical" studies, and the important issue is not whether differences existed, but whether these differences were critical. If we accept the validity of our results, then in retrospect the two studies must have differed in significant ways. But since we were unable to specify these critical differences in advance, we are left with no alternative but to consider our studies inconclusive regarding the effects of the experimental manipulations on ESP performance. The difficulty of specifying what differences may have been critical is a direct result of our lack of knowledge and theory about psi. Given a better defined theory of psi, it would have been possible to know in advance which variables were critical and which not. We suggest that the present experimental approach to psi may help in developing the necessary theory. Consider the following: It has been suggested that the tone of the subject-experimenter interaction may influence psi (Schmeidler, 1969). Consistent with this possibility, large and statistically significant differences in subjects' self-ratings and ratings of the experimenter occurred between the first and second experiment. Perhaps, then, our experimental design confounded a critical variable. By implementing the necessary controls in future research, this problem could be reduced. Videotaping the experimental instructions, for example, would help control for experimenter-subject interaction differences within the sessions of one experiment and across any attempts at replication. On the basis of the results of such experiments, the necessary adjustments could be rendered to the developing theory of psi. Experimental social psychologists have had a long history of grappling with evanescent, apparently intractable phenomena. Over time and with constant refining of experimental designs, most of these "intractable" phenomena have yielded. We are confident that an experimental approach to psi, in the spirit of the present research, will likewise advance our understanding of this phenomenon. REFERENCES AJZEN, I. Additudinal vs normative m e s s a g e s : A n investigation of the differential effects of persuasive c o m m u n i c a t i o n s on behavior. Sociometry, 1971,34, 2 6 3 - 2 8 0 . AJZEN, 1., 8r FISHBEIN, U . T h e prediction of behavioral intentions in a choice situation. Journal of Experimental Social Psychology, 1969, 5, 4 0 0 - 4 1 6 . AJZEN, I., & FISHBEIN, M. T h e prediction of behavior from attitudinal and normative variables. Journal of Experimental Social Psychology, 1970, 6, 4 6 6 - 4 8 7 .
BELIEF, EVALUATION, AND ESP
179
AJZEN, I., & FISHBEIN, M. Attitudes and normative beliefs as factors influencing behavioral intentions. Journal of Personality attd Social Psychology, 1972, 21, 1-9. AJZEN, I., & FISHBEIN, M. Attitudinal and normative variables as predictors of specific behaviors. Journal of Personality and Social Psychology, 1973, 27, 41-57. BEVAN, J. M. The relation of attitude to success in ESP scoring. Journal r Parapsychology, 1947, I l, 296-309. BHADRA, B. H. The relationship of test scores to belief in ESP. Join'hal (~f" Parapsychology, 1966, 30, I- 17. CAMPBELL, D. T. Conformity in psychology's theories of acquired behavioral dispositions. In I. A. Berg and B. M. Bass (Eds.), Conformity and deviation. New York: Harper, 1961. Pp. 101-142. COLLINS, B. E., ASHMORE, R. D., HORNBECK, F. W., t~ WHITNEY, R. E. Studies in forced compliance: XIII and XV. In search of a dissonance-producing forced compliance paradigm. Representative Research in Social Psychology, 1970, 1, 11-23. COOPER, G. W. A further study of the relation of attitude to success in ESP scoring. Journal ~)["Parapsychology, 1951, 15, 139-145. EILBERT, L., ~r SCHMEIDLER, G. R. A study of certain psychological factors in relation to ESP performance. Journal of Parapsychology, 1950, 14, 53-74. FISHBEIN, M., & RAVEN, B. H. The AB scales: An operational definition of belief and attitude. Human Relations, 1962, 15, 35-44. LAPIERE, R. T. Attitudes vs actions. Social Forces, 1934, 13, 230-251. McGUIRE, W. J. The nature of attitudes and attitude change. In G. Lindzey and E. Aronson (Eds.), Handbook of social psychology, Vo]. 3, Reading, Mass.: AddisonWesley, 1969. P p 136-314. MtJsso, J. R. ESP experiments with primary school children. Journal of Parapsychoh~gy, 1965, 29, 115-121. ORNE, M. T. On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American Psychologist, 1962, 17, 776-783. PALMER, J. Scoring in ESP tests as a function of belief in ESP. Part I: The sheep-goat effect. Journal <)f the American Society fi)r Psychical Research, 1971, 65, 373-408. PALMER, J. Scoring in ESP tests as a function of belief in ESP. Part I1." Beyond the sheepgoat effect. Journal of the American Society jor Psychical Research, 1972, 66, 1-26. RVZL, M. Precognition scoring and attitude. Journal (~f Parapsychology, 1968, 32, 183-189. SCHMEIDLER, G. R. Predicting good and bad scores in a clairvoyance experiment: A final report. Journal of the American Society Jbr Psychical Research, 1943, 37, 210-227. SCHMEIDLER, G. R. Extrasensory perception. New York: Atherton Press, 1969. SILVERMAN, I. Role-related behavior of subjects in laboratory studies of attitude change. Journal ~)f Personality and Social Psychology, 1968, 8, 343-348. STANFORD, R. G. A further study of high- versus low-scoring sheep. JouHtal of Parapsychology, 1965, 29, 141-148. VAN DE CASTLE, R. L., & WHITE, R. A report on a sentence completion form of sheepgoat attitude scale. Journal t~fParapsychology, 1955, 19, 171-179. WOODRUFF, J. L., & DALE, L. Subject and experimenter attitudes in relation to ESP scoring. Journal of the American Society fi~r Psychical Research, 1950, 44, 87-112.