The Psychological Record, 2005, 55, 115-123 USE OF AN ESP COVER STORY FACILITATES REINFORCEMENT WITHOUT AWARENESS LEWIS A. BIZO and NICOLA SWEENEY University of Southampton Participants were exposed to 3 conditions in a betweengroups design. Participants were told the experiment was about extrasensory perception and were asked to select the word from a pair of words that they thought the experimenter was thinking about. In 1 condition selections of the word with a double letter were reinforced with positive verbal feedback, and in another condition selections of nondouble letters were reinforced. The word selections of a control group were reinforced according to a random and predetermined order. By the end of the experimental session, participants who were reinforced for selecting a particular type of word, double letter or nondouble letter, selected that word significantly more than the alternative that was not associated with reinforcement and those participants in the control group selected the double-letter and nondouble-ietter words equally often. In a postexperimental questionnaire none of the participants reported an awareness of the contingency. The present experiment replicates and extends previous work and confirms that participants' behavior can be modified by reinforcement without the participants' conscious awareness of the contingency. Recently, Lieberman and colleagues (Lieberman, Connell, & Moos, 1998; Lieberman, Sunnucks, & Kirk, 1998) reported the results of a series of experiments that showed that aspects of participants' behavior could be reinforced despite their inability to articulate what aspect of their behavior was being reinforced. Lieberman, Sunnucks, et a'l. (1998) reinforced louder responses and Lieberman, Connell, et al. (1998) differentially reinforced responses to double-letter words in a two-choice task. This naivety to the prevailing contingency was facilitated by the use of a cover story: Lieberman, Connell, et al. (1998) told participants that they were participating in an experiment on extrasensory perception (ESP). Elsewhere, certain aspects of human behavior have been brought under the control of schedules of reinforcement when instructions focus attention away from the controlling contingency (e.g., Svartdal, 1989, 1991). Correspondence concerning this article should be addressed to Lewis A. Bizo, School of Psychology, University of Southampton, Southampton, S017 1 BJ, United Kingdom. (Email: Lewisb@soton.ac.uk).
116 SIZO AND SWEENEY Much of the research on learning without conscious awareness has been focused on complex rules or abstract tasks learned through a phenomenon defined as implicit learning by Reber (1967). Reber (1989) focused on complex grammar tasks in which participants appear to unconsciously learn complex grammar sequences in order to choose the appropriate responses. However, when asked what hypotheses they used to make their decision they either appeared unsure or suggested an alternative hypothesis unrelated to the task being tested. This inability to articulate the prevailing contingency has also been documented on simple schedules of reinforcement. For example, Bruner and Revusky (1961) observed the development of idiosyncratic stereotyped responding during the waiting phase of a differential reinforcement of low rates (DRL) schedule with human subjects. In postexperimental interviews, however, all participants stated that they thought reinforcement was contingent upon patterns of responding, and none realized the temporal contingency. There are numerous examples of human insensitivity to schedules of reinforcement, where despite changes in the contingency of reinforcement, relevant aspects of behavior do not change (for reviews, see: Kollins, Newland, & Critchfield, 1997; Madden, Chase, & Joyce, 1998). However, manipulations of the experimental instructions can facilitate schedule control (e.g., Galizio, 1979; Hackenberg & Joker, 1994). Instructions can constrain responding in such as way as to facilitate contact with any arranged contingency (Hayes, 1989). For example, the use of a task that all participants interpret similarly appears to attenuate the unpredictable effects on performance produced by self-instruction (Bizo, Remington, D'Souza, Heighway, & Baston, 2002). In the present experiment we intended to extend the work of Lieberman, Connell, et al. (1998) on reinforcing word class. The use of a cover story that encourages participants to believe the experiment is designed to test some aspect of ESP is useful for several reasons: Bem and Honorton (1994) have noted that lay persons are more accepting of the idea of an alternative sensory system, such as ESP, thus participants in the studies by Lieberman et al. were more likely to have accepted the notion that the task they were completing actually tested ESP; Bem and Honorton also argue that forced-choice procedures, as used by Lieberman, Sunnucks, et al. (1998) and by Lieberman, Connell, et al. (1998) in their experiments, fail to capture the circumstance that characterize reported instances of Psi; and most importantly Bem and Honorton (1994) noted that ESP believers' performance on forced-choice procedures is no better or worse than nonbelievers'. The ESP task, therefore, should act as a suitable distractor and should attenuate participants testing of alternative hypotheses. The main aims of the present experiment were to test an individual's ability to learn to select a particular type of word and whether or not participants would be able to articulate the contingency. That is, would participants learn a simple word task (i.e., picking the words containing a
REINFORCEMENT WITHOUT AWARENESS 117 double letter) through simple reinforcement without being consciously aware of what they were doing. Participants in the study by Lieberman, Connell, et al. received reinforcement for selecting a double-letter word from a choice of two words in a forced choice test. We have adapted the procedure of Lieberman, Connell, et al. (1998) by including a common baseline of 10 rather than 40 trials when all participants receive the same feedback regardless of which word they chose. The experiment makes use of three conditions arranged as a between-groups design. Selections of words with double letters will be reinforced in one condition, and selections of words with no double letters will be reinforced in another condition. We reinforced both double-letter and nondouble-ietter choices whereas Lieberman, Connell, et al. (1998) only reinforced double-letter choices. We expected that by the end of the experimental session participants who were reinforced fo r selecting a particular type of word, double letter or nondouble letter, would select that word type more than the alternative that was not associated with reinforcement, and that those participants in the control group would select the double-letter word about 50% of the time. Method Participants The participants were 30 undergraduate students from the University of Southampton who participated for partial course credit. Twenty two were female and 8 were male, the average age of the participants was 22 years (range from 18-43 years). Participants were recruited via a sign-up board for psychology students, and the sign-up sheet for the experiment asked for volunteers to participate in an experiment on extrasensory perception (ESP). Apparatus The participants sat on one side of a table with a white screen positioned on it (.75 m high x 1 m long) and the experimenter sat on the other side of the table out of sight of the participant. A 2-cm gap between the bottom of the screen and the table allowed the experimenter to present the choice stimuli to the participant. The screen was used to prevent nonverbal cues during the experiment. One hundred and fifty pairs of cards were used. Double-letter words printed on half the cards in 12 point font and nondouble-ietter words were printed on the other half (e.g., thread, llama). A 10-question postexperimental questionnaire, based on that used by Lieberman, Sunnucks, et al. (1998), was used to assess participants' awareness of the schedule, use of rules, and whether they had generated and tested any hypotheses. Procedure At the start of the experiment participants were randomly assigned to one of three conditions. In the double letter (DL) condition the selection of
118 SIZO AND SWEENEY the word that contained a double letter was reinforced. In the no double letter (NDL) condition, selection of the word that did not contain a double letter was reinforced. In the control condition, reinforcer delivery was unrelated to the word selected by participants and was delivered according to a random and predetermined order. All participants in the control condition received the same sequence of reinforcers. Participants were asked before they started the experiment if they had spoken to anyone else about the experiment-none reported that they had. Once settled comfortably in the room, they read the standardized experimental instructions which informed the participants: that they were participating in an experiment on extrasensory perception (ESP); that they were to select the card, from a choice of two, that the experimenter was thinking of and to read it out aloud; that the experimenter would respond "yes" if they had selected the correct card or say nothing at all if they selected the incorrect card. Baseline. Baseline testing involved presenting participants with 10 card pairs, two at a time. Each card had one word on it and one of the words contained a double letter. Participants were asked to select the card they thought the experimenter had in mind. During baseline testing, selections were reinforced randomly according to a predetermined order such that participants were told "yes" they had selected the correct card on 5 of the 10 trials. Test phase. Immediately following the baseline testing, the test phase began which involved the presentation of 140 card pairs. Participants were not told the test phase had begun. Once participants had completed all 140 trials they were asked to complete the postexperimental questionnaire. After all participants had been tested, they were fully debriefed via e-mail about the use of deception in the experiment and the nature of the hypothesis. They were also provided with a summary of the results. Results The mean number of double-letter responses for each of the three groups over the 15 blocks of 10 trials is shown in Figure 1. The three groups showed clear differences in the mean number of double-letter responses they gave. A one-way analysis of variance (ANOVA) revealed that the number of double-letter words selected did not differ significantly between the three groups during the baseline block of trials, F(2, 27) = 1.43, P = 0.25, but were significantly different by the last block of trials, F (2, 27) = 37.40, p < 0.05. The number of double-letter words selected by participants in the DL condition increased significantly from a mean of 4.5 words during the baseline block of 10 trials to 7.2 words by the 15th block of 10 trials, ~9) = -5.45, P < 0.05, and only 1 participant in this group choose the same number of double-letter words at the end of the experiment as they did in baseline and did not show an increase in double-letter choices. For participants in the NDL condition the number of
REINFORCEMENT WITHOUT AWARENESS 119 double-letter words selected decreased significantly from a mean of 5.2 words during the baseline block of 10 trials to 2.7 words by the 15th block of 10 trials, ~9) = 5.51, P < 0.05, and only 1 participant in this group chose the same number of double-letter words at the end of the experiment as they did in baseline and did not show a decrease in double-letter choices. For the control group, the number of double-letter words selected showed a nonsignificant increase from a mean of 4.3 words during the baseline block of 10 trials to 5.1 words by the 15th block of 10 trials, ~9) = -1.92, P = 0.09. en 10 en Control c:: 0 9 -&- Double Letter 0. en 8 ---'V- Non-Double Letter 0:::... :j:::...j :0 :::I 0 0-0....0 E 7 6 5 4 3 2 :::I Z c:: 1 ro ::a: 0 0 3 6 9 12 15 Block Number Figure 1. Mean number of double-letter responses for each of the three experimental conditions over 15 blocks of 10 pairs. The dashed horizontal line shows chance performance. Table 1 Percentage of Participants Who Responded Positively to Postexperimental Questions Percentage of positive Question responses (%) Do you believe in ESP? 70 Do you feel you concentrated to the best of your ability in order to pick up 60 the signals the experimenter was passing on? Did you feel that you successfully received any ESP signals? 33.3 Did you feel that any other variables were aiding or abetting your ability to pick up signals? 26.7
120 SIZO AND SWEENEY Postexperimental questionnaire. The ESP cover story appeared to be successful, with the analysis of the questionnaire data (see Table 1) revealing that 70% of participants reported that they believed in ESP, and 33% purported to have received signals during the experiment. The three groups did not differ greatly in terms of the numbers of participants that reported that they believed in ESP; 8 in the DL condition, 7 in the NDL, and 6 participants in the control condition reported that they believed in ESP. The three groups did differ slightly in terms of the numbers of participants that reported that they had received ESP signals; 5 in the DL condition, 3 in the NDL condition, and 2 participants in the control condition reported that they had received ESP signals. Whether a participant believed in ESP or not did not affect the extent to which their choice of a double-letter word was affected by the prevailing reinforcement contingency susceptibility-the belief of participants in ESP was not significantly correlated with absolute percentage change in double-word choices for participants in the DL and NDL conditions (rs = 0.131, df = 20, P =.582), and the belief of participants that they had received an ESP signal was not significantly correlated with absolute percentage change in double-word choices for participants in the DL and NDL conditions (r s = 0.142, df = 20, P =.549). None of the participants reported any awareness of the true nature of the hypothesis being tested (see Table 2). Some participants composed hypotheses but none of these corresponded to the contingency that was in effect, although 1 participant in the control group reported thinking that the words were being chosen randomly. Three participants suggested word length which was related to the presence of a double letter. The double-letter words were slightly longer than nondouble-ietter words, M = 6.0 and 5.3 letters, respectively. The nondouble-ietter words were the same length or longer than the double-letter words on 47% of the trials and shorter on the remaining 53% of trials. It is possible that word length might have functioned as a discriminative stimulus and thus a vehicle for self-generated hypotheses for at least 3 participants, however, word length was not as predictive a discriminative stimulus as the presence or absence of a double letter that was predictive of a reinforcer on 100% of trials for the experimental groups (DL and NDL groups). Discussion The present experiment replicates and extends previous work and confirms that participants' behavior can be modified by reinforcement without participants' conscious awareness of the contingency. The change in performance of the DL and NDL groups replicates the findings of Lieberman, Sunnucks, et al. (1998) and Lieberman, Connell, et al. (1998). The results are consistent with the hypothesis that learning can occur without conscious awareness. In the DL condition the mean number of times the participants chose a word with a double letter increased significantly over the course of testing; whereas in the NDL
REINFORCEMENT WITHOUT AWARENESS 121 condition, the mean number of times the participants chose a word with a double letter decreased significantly over the course of testing. Participants in the control group showed no significant change in the mean rate of double-letter responses over the 15 trials. The reinforcement of selection of the double/nondouble lettered word increased the likelihood of that response in the absence of explicit instruction to do so. The significant difference between the three conditions suggests that participants' behavior was controlled by the prevailing contingency. The written transcripts (Table 2) show that not a single participant articulated that they were being reinforced for the type of word they chose, even though the behavior of participants on the experimental condition was influenced by the prevailing contingency. The absence of Table 2 Transcript of Responses to Question "Did you compose any hypotheses for choosing the words?" Double-letter Group Participant 1 No 2 Words that have set orders 3 Most of the time the harder word of the two was right 4 No 5 No 6 No 7 Word length?! 8 I think she was forcing the words through priming 9 No 10 Certain meanings Nondouble Letter Group 11 Was there some animal theme?! 12 13 No 14 No 15 No 16 17 Word order 18 No 19 Stronger images maybe (but that soon went) 20 I thought that they might have been chosen randomly Control Group 21 The most interesting words in my opinion I thought would be chosen by the experimenter 22 Couldn't quite get the word significance- but could work out some of the pattem. 23 No, none at all 24 No 25 No 26 27 28 No 29 Not really, it was just random selection 30 Yes, the hand the word was in, the spelling of the words, length of words,
122 81Z0 AND SWEENEY any significant correlation between the results of the postexperimental questionnaire and the number of double-letter responses is consistent with results of Bem and Honorton (1994) who found that belief in ESP had no effect on performance in matching tasks of this type. Although it is unlikely that ESP exists and therefore unlikely that any one individual's psychic ability could have biased the results, it is possible that strength of belief in the cover story could have had an effect, if those participants were more susceptible to the subtle effects of the contingency employed. Although our results did not show significant correlation between belief and percentage change in performance, our index of belief in ESP did not distribute across a continuum, and only registered a "yes" or "no" response. If some participants were more sensitive than others to the effects of the schedule of reinforcement then this could have been examined more successfully if a Likert scale had been used to rate the beliefs' effects on participants in the postexperimental questionnaire. It is worth commenting on the high percentage of participants who responded on the postexperimental questionnaire that they believed in ESP (70%), and a third of participants reported that they had successfully received ESP signals. Our experiment probably appealed to individuals with an interest in ESP because we explicitly recruited for volunteers to participate in an experiment on ESP-we would be surprised if the prevalence of this belief in ESP was held by a wider population of psychology students. However, further research could look at the sensitivity of indi\liduals who believe in ESP on these types of tasks. A similar technique was used by Svartdal (1989) to test how meaningful participants had found his experiment. It is also possible in this experiment that participants that did not believe in ESP would have discounted the experimental instructions and perhaps been more likely to formulate and test alternative hypotheses which could have blocked the effect of the reinforcement contingency. Svartdal (1989) suggested that improbable or impossible verbal guidance increases the participant's sensitivity to nonverbal contingencies; this also seems to be confirmed by the lack of correlation in these results. It is likely these types of "situated tasks" minimize the unpredictable effects of self-instruction by focusing participants' attention on a clearly articulated task (e.g., Bizo et ai., 2002; Lieberman, Connell, et ai., 1998; Lieberman, Sunnucks, et ai., 1998 Svartdal, 1989, 1991). The ESP cover story can be assumed to be improbable and virtually impossible, and perhaps therefore it renders the participants more sensitive to the reinforcement contingency to choose either the double letter or nondouble letter. References BEM, D. J., & HONORTON, C. (1994). Does psi exist? Reliable evidence for an anomalous process of information transfer. Psychological Bulletin, 115, 4-18.
REINFORCEMENT WITHOUT AWARENESS 123 BIZO, L. A., REMINGTON, B., D'SOUZA, L. S., HEIGHWAY, S. K., & BASTON, C. (2002). Human variable ratio performance. Learning and Motivation, 33, 411-432. BRUNER, A., & REVUSKY, S. H. (1961). Collateral behavior in humans. Journal of the Experimental Analysis of Behavior, 4, 349-350. GALIZIO, M. (1979). Contingency-shaped and rule-governed behavior: Instructional control of human loss avoidance. Journal of the Experimental Analysis of Behavior, 31,53-70. HACKENBERG, T. D., & JOKER, V. R. (1994). Instructional versus schedule control of humans' choices in situations of diminishing returns. Journal of the Experimental Analysis of Behavior, 62, 367-3133. HAYES, S. C. (1989). Rule governed behavior: Cognition, contingencies and constructional control. New York: Plenum. KOLLlNS, S. H., NEWLAND, C., & CRITCHFIELD, T. S. (1997). Human sensitivity to reinforcement in operant choice: How much do consequences matter? Psychonomic Bulletin & Review, 4, 208-:220. LIEBERMAN, D. A., CONNELL, G. L., & MOOS, H. F. T. (1998). Reinforcement without awareness: II. Word class. The Quarterly Journal of Experimental Psychology, 51 B, 317-335. LIEBERMAN, D. A., SUNNUCKS, W. L., & KIRK, J. D. J. (1998). Reinforcement without awareness: I. Voice level. The Quarterly Journal of Experimental Psychology, 51 B, 301-316. MADDEN, G. J., CHASE, P. N., & JOYCE, J. H. (1998). Making sense of sensitivity in the human operant literature. The Behavior Analyst, 21, 1-12. REBER, A. S. (1967). Implicit learning of artificial grammars. Journal of Verbal Learning and Verbal Behavior, 5, 855-863. REBER, A. S. (1989). Implicit learning and tacit knowledge. Journal of Experimental Psychology: General, 118, 219-235. SVARTDAL, F. (1989). Shaping of rule-governed behaviour. Scandinavian Journal of Psychology, 30, 304-314. SVARTDAL, F. (1991). Operant modulation of low-level attributes of rule-governed behavior by nonverbal contingencies. Learning and Motivation, 22, 406-420.