{NCEE‘WVE AS A METiGATGR OF THE EXTENT T0 WHEEH GNE'S ATTITUDES ARE ENFERRED FRGM QNE’S BEI‘iA‘JiOR: STUDEES IN THE THESRY 0F SELF-PERCEPTIOE‘E Thesis for the Degree of Ph. D. MECHIGAN STATE UNIVERSITY R. SAMUEL MEHRLEY 1970 IIHH 1 ll Wilmmflflwlmm‘ml ‘ .... 3 1293 10747 6685 1;,2’1»: Y “‘3 . Michigan State University J {HES}? This is to certify that the thesis entitled INCENTIVE AS A MITIGATOR OF THE EXTENT TO WHICH ONE'S ATTITUDES ARE INFERRED FROM ONE'S BEHAVIOR: STUDIES IN THE THEORY OF SELF-PERCEPTION presented by R. Samuel Mehrley has been accepted towards fulfillment of the requirements for Ph.D. degree in Communication . Date January 15; [97] 0-169 “a... ~_~W- r - V 1:.1'111 [I‘ll I’ll li‘lll ABSTRACT INCENTIVE AS A MITIGATOR OF THE EXTENT TO WHICH ONE'S ATTITUDES ARE INFERRED FROM ONE'S BEHAVIOR: STUDIES IN THE THEORY OF SELF-PERCEPTION BY R. Samuel Mehrley Two separate studies investigated the self-perception hypothesis that an individual's attitude statements may be predicted and controlled by manipulating his overt behavior and the stimulus conditions under which it occurs. In Experiment I gs administered positive or negative feedback to an unknown Speaker under conditions of high or low in- centive. The primary dependent variable was ratings of speaking effectiveness. Experiment II was a replication of Experiment I with the exception that speech quality (high and low) was manipulated. Experiment I provided strong support for the hypothesis that an individual infers his own attitudes from his own behavior while Experiment II provided only a modicum of support. Neither experiment supported the hypothesis that incentive mitigates the extent to which one's attitudes are inferred from one’s R. Samuel Mehrley behavior. In fact, in EXperiment II reinforcement effects occurred in belief-consistent conditions while self- perception effects occurred in belief-inconsistent conditions. INCENTIVE AS A MITIGATOR OF THE EXTENT TO WHICH ONE'S ATTITUDES ARE INFERRED FROM ONE'S BEHAVIOR: STUDIES IN THE THEORY OF SELF-PERCEPTION By 62‘ R? Samuel Mehrley A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Communication 1970 Accepted by the faculty of the Department of Communi- cation, College of Communication Arts, Michigan State Uni- versity, in partial fulfillment of the requirements for the Doctor of PhilOSOphy degree. WWW)! D.rector of Thesis ’Vf / ' I «I Guidance Committee: . __' ‘g‘A/Al A [1/], Chairman - /,. LM€I 222': . Q2 , ACKNOWLEDGMENTS Intellectual debts can only be acknowledged, never repaid. Many scholars and friends (not mutually exclusive categories) whom I can never thank adequately have con- tributed ideas and encouragement at various stages of my development. Professor Forrest D. Tucker of the University of Southern Mississippi first whetted my appetite for the study of communication. Professor James C. McCroskey, now of Illinois State University, provided the prodding, en- couragement, and insight needed to guide me through the first two years of graduate school. Professor Gerald R. Miller has made contributions whose worth cannot be weighed. His writing and teaching have permeated my own thinking beyond recognition. Three other individuals--Professors William Crano, David Ralph, and Lawrence Sarbaugh--served as conscientious members of my guidance committee. They were always willing and able to give needed assistance. The price a wife pays while her husband is in gradu- ate school is high. She must tolerate his obsession and his unpredictable ups and downs. I express my gratitude to Sara for her ability to cope. ii Finally, I express my appreciation to my earliest and most influential teachers, my parents, Robert Mehrley and Freda Mehrley, by dedicating this paper to them. iii TABLE OF CONTENTS Chapter I. INTRODUCTION . The Problem. Relevant Research. Rationale and Hypotheses II. EXPERIMENT I . Overview. . Subjects. . Procedures . Results . . Discussion . III. EXPERIMENT II. Overview. . Rationale and Subjects. . Procedures . Results . . Discussion . IV. DISCUSSION. . Implications Summary . . REFERENCES . . . . APPENDICES Appendix A. Pretest Attitude Questionnaire, Experiment Hypotheses for Further and Experiment II. Research B. Experimental Speech, EXperiment I . iv Page 21 21 22 22 30 50 55 55 55 57 58 66 86 96 100 102 104 108 Ill Chapter Page Appendix C. Posttest Attitude Questionnaire, Experiment I and Experiment II . . . . 115 D. Control Posttest Attitude Questionnaire, Experiment I . . . . . . . . . . 124 E. Experimental Speeches, Experiment II. . . 133 10. LIST OF TABLES Page Means, Standard Deviations, and Analysis of Variance of Incentive Ratings . . . . . 31 Eftest Comparisons of Incentive Ratings of Experimental Groups . . . . . . . . 32 Percentage and Frequency of Subjects Accurately Perceiving the Feedback Condition. . . . . . . . . . . . 32 Mean Ratings and Standard Deviations for Experimental and Control Groups Within the Four Speech Rating Dimensions . . . . . 34 Summary Table of Analysis of Variance of Speech Ratings of Experimental and Control Groups. . . . . . . . . . . . . 35 Analysis of Differences of Speech Ratings Between Each Experimental Group and contr01 O O O O O O O O O O O O 36 Summary Table of Two-Factor Analysis of Variance of Speech Ratings . . . . . . 38 Estest Comparisons of Mean Speech Ratings of Experimental Groups . . . . . . . 39 Mean Ratings and Standard Deviations for Experimental and Control Groups Within the Three Credibility Dimensions . . . . . 42 Summary Table of Analysis of Variance of Credibility Ratings of Experimental and Control Groups . . . . . . . . . . 42 vi Table Page 11. Summary Table of Two-Factor Analysis of Vari- ance of Credibility Ratings . . . . . . 43 12. Estest Comparisons of Mean Credibility Ratings of Experimental Groups . . . . . 45 13. Means, Standard Deviations, and Analysis of Variance of Pretest Attitude Scores for EXperimental and Control Groups. . . . . 47 14. Attitude Scores, Amount of Change, and Analysis of Variance of Attitude Change Scores of Experimental and Control Groups . 48 15. Summary Table of Two—Factor Analysis of Vari— ance of Attitude Change Scores of Experimental Groups. . . . . . . . . 49 16. tetest Comparisons of Mean Attitude Change Scores of Experimental Groups . . . . . 49 17. Pretest Means, Standard Deviations, and Eftest Comparisons of the High Quality Speech Versus the Low Quality Speech Within the Four Speech Rating Dimensions . . . . . 62 18. Means, Standard Deviations, and Analysis of Variance of Incentive Ratings . . . . . 68 19. Estest Comparisons of Mean Incentive Ratings of Experimental Groups. . . . . . . . 69 20. Percentage and Frequency of Subjects Accur- ately Perceiving the Feedback Condition . . 7O 21. Mean Ratings and Standard Deviations for Experimental Groups Within the Four Speech Rating Dimensions . . . . . . . . . 72 22. Summary Table of Three-Factor Analysis of Variance of Speech Ratings . . . . . . 73 23. Eftest Comparisons of Mean Speech Ratings of Experimental Groups. . . . . . . . . 76 24. Mean Ratings and Standard Deviations for Experimental Groups Within the Three Credibility Dimensions. . . . . . . . 77 vii {.I Ilti r |II|. {In Table 25. 26. 27. 28. 29. 30. Page Summary Table of Three—Factor Analysis of Variance of Credibility Ratings . . . . 79 Eftest Comparisons of Mean Credibility Ratings of Experimental Groups . . . . 81 Means, Standard Deviations, and Analysis of Variance of Pretest Attitude Scores for Experimental Groups . . . . . . . . 83 Mean Pretest, Posttest, and Attitude Change Scores for Experimental Groups . . . . 84 Summary Table of Three-Factor Analysis of Variance of Attitude Change Scores of Experimental Groups . . . . . . . . 85 Estest Comparisons of Mean Attitude Change Scores of Experimental Groups. . . . . 87 viii CHAPTER I INTRODUCTION The Problem Arguing from a radical Skinnerian position, Bem (1965, 1966, 1967, 1968, 1970a, 1970b) has proposed that individuals come to "know" their own attitudes and other internal states partially by inferring them from obser- vations of their own overt behavior and the circumstances in which it occurs. Thus, the individual is functionally in the same position as an outside observer of his be— havior, an observer who, necessarily, must rely upon those same external cues to infer the individual's inner states. For example, an observer judges another's atti- tude on the basis of his behavior and the stimulus con- ditions under which it occurs. Similarly, an individual frequently judges his own attitude from his own behavior, taking account of the conditions under which it occurs. In short, Bem's interpersonal theory of self-perception suggests that the process of arriving at inferences about .the self are identical with those used in drawing infer- ences about others. It is this assertion, that self- perception is limited to the same set of public cues used in person perception, which constitutes Bem's distinct theoretical contribution. Despite the plausibility and potential importance of Bem's notion, there is a paucity of direct empirical sup- port for the basic theoretical premises. The studies re- ported below attempt to extend and refine the current level of understanding regarding Bem's theory of self-perception by examining the efficacy of applying it to communication behavior. In particular, the studies deal with the general hypothesis that an individual's belief and attitude state- ments may be predicted and controlled by manipulating his overt behavior and the stimulus conditions under which it occurs. A review of relevant research clarifies the relationship between these variables and provides a basis for the hypotheses which were investigated. Relevant Research This section reviews research which is relevant to the basic theoretical premises of Bem's self-perception theory. In addition, Bem's analysis of dissonance theory experiments is presented. Empirical Support for the Basic Theoretical Premises Bem's basic premise is that "In identifying his own internal states, an individual partially relies on the same external cues that others use when they infer his internal states" (1970, p. 50). Bem identifies three im- portant sources of such cues: (1) those external cues residing in the social or physical situation in which the individual is placed; (2) those external cues residing in the individual's own behavior; and (3) those external cues residing in the circumstances that appear to be responsible for the individual's behavior. While not conducted as tests of self-perception theory, there are several studies that lend support to the notion that individuals use cues residing in the social or physical situation in which they are placed to infer their own internal states. Schachter and Singer (1962) investi- gated the combined effects of internal and external cues on the subject's perceptions of his own emotionality. Subjects were injected with a drug which provokes the in- ternal physiological reactions that accompany strong emotions. The subjects, however, were not informed of the true effects of the injection. They were placed in a waiting room with a confederate who pretended to be another subject. In one condition the confederate behaved in an angry manner while in a second condition he behaved in a happy manner. A control group received a placebo instead of the drug. The results revealed that subjects who had been injected with the placebo perceived themselves to be relatively unemotional after the session with the confederate regardless of how the confederate had behaved. Subjects who had been injected with the drug felt quite differently. Those who had been with the angry confederate described their mood as one of anger while those who had been with the happy confederate felt slightly "eupheric." Finally, drugged subjects did not describe themselves as emotional if they knew ahead of time what reactions to expect from the drug. Valins (1966) conducted a study in which subjects were shown a series of slide pictures of seminude females and at the same time heard what they believed to be the amplified sounds of their hearts beating. Variations in rate changes of the sounds were coordinated with presen- tation of the slides. Thus, in one condition, each subject heard his heart rate increase soon after the presentation of each of the slides (and this occurred on both of two successive exposures to the series of slides), but he noted no such increase after five other pictures. In a second condition, subjects heard decreases in rate rather than increases. The effect of the covariation procedure was that subjects rated as more attractive the pictures with which the changes had been associated and more frequently chose them as remuneration for their participation in the experiment. Schachter and Gross (1968) based a study on the assumption that overweight individuals may eat too much or too often because they do not rely upon internal stomach cues to tell themselves that they are hungry, but instead rely primarily upon external circumstances. The procedures consisted of requiring both obese and normal subjects to sit quietly in a room for 30 minutes with nothing to do. The room contained a clock which was rigged to run fast for some subjects and slow for others. The experimenter left the room when the clock read 5:05 P.M. When he re- turned a half hour later, the clock read 6:05 for some subjects but only 5:20 for others. Each subject was then allowed to have a snack. The results revealed that obese subjects ate more when the clock read 6:05 than when it only read 5:20. Conversely, normal-weight subjects ate less when the clock read 6:05; a few of them commented that they did not want to spoil their appetites for dinner. This study suggests that obese individuals rely less upon internal cues than normal individuals. This sug- gestion leads to the conclusion that obese individuals should actually be better able to tolerate food depri- vation than normal individuals if the external cues are removed. Goldman, Jaffa, and Schachter (1968) investi- gated this possibility. They discovered that obese Jews were more likely to fast on Yom Kippur, the Jewish Day of Atonement, than were normal-weight Jews. Furthermore, the more time obese Jews spent in the synagogue (away from external food cues), the easier it was for them to fast. Conversely, the normal-weight Jews carried their (internal) hunger cues into the synagogue with them. For them, being in the synagogue had little to do with the difficulty of the fast. Bem asserts that to an observer the most important clues to an individual's inner states are found in his be- havior. Accordingly, Bem's theory of self-perception pre- dicts that an individual also infers his own internal states by observing his own overt behavior. In an effort to pro- vide direct empirical support for this prediction, Bandler, Madaras, and Bem (1968) asked subjects to press a button in order to escape electric shock. The subjects were unaware that all the shocks were of equal intensity. Prior to each shock, the experimenter told the subject whether he pre- ferred the subject to escape it or to endure it but that the final choice was his. In this way, the experimenter was able to control which shocks the subject terminated and which he endured, even though he saw himself as having a choice in the matter. After each shock, the subject rated the degree of discomfort from it. The results re- vealed that subjects rated the shocks as significantly more uncomfortable when they escaped them than they did when they endured them. This was true even though the endured shocks were necessarily longer than the escaped shocks. Kiesler, Nisbett, and Zanna (1969) led subjects to believe that they were going to deliver arguments against air pollution to passers-by in the street. At the same time, confederates were asked to argue for promoting auto safety. The confederate stated that he was willing to argue in favor of auto safety because (a) he believed strongly in it (belief-relevant condition), or (b) the experiment was scientifically valuable (belief-irrelevant condition). Belief-relevant subjects were found to be more opposed to air pollution than either belief-irrelevant subjects or control subjects who were not committed to the behavior. The results support Bem's hypothesis that people infer their beliefs from their behavior and suggest that such a process occurs only when there are cues implying that behavior is relevant to belief. While not conducted as tests of self-perception theory, there are at least two other studies that lend themselves to a Bemian interpretation. Because of their relevance to the studies reported below, they merit a detailed examination. Amato and Ostermeier (1967) investigated the oper- ation of feedback in a public speaking situation. The audience consisted of 22 students enrolled in a funda- mentals public speaking course. Three other students en- rolled in a fundamentals public speaking course, unknown to the audience, served as subject-speakers. Each audience member was given three index cards to use in presenting feedback during the oral presentation of each speaker. The audience was instructed to assume a predetermined response-role for each speaker. In one condition the audience was instructed to respond favorably to the speaker by exhibiting a WHITE card. In the other conditions they were instructed to indicate an unfavorable or neutral response by exhibiting either a RED or GREEN card re- spectively. These audience response-role procedures were followed for each of the three speakers during the first speaking session. The speakers were given the same in- structions concerning the meaning of the colored cards. At the conclusion of each speech, the audience completed a questionnaire in which they were asked to evaluate the speaker on six speaking characteristics. A second set of instructions was given the audience prior to the second round of speeches. In this set only two response-roles were permitted, either unfavorable or neutral but not favorable. Again the questionnaires were administered following each speech. Results were reported to be sup- portive of the researchers' hypothesis: that unfavorable audience feedback produces a deterioration in speaker delivery; specifically, nervousness, bodily movement, eye contact, and fluency. Combs and Miller (1968) argued for a more parsi- monious explanation of Amato and Ostermeier's findings; namely, that the audience members rated the speakers con- sistent with the response-role they were asked to perform. They hypothesized that subjects administering positive feedback to a speaker will rate him more favorably than subjects administering neutral feedback, and that subjects administering negative feedback to a speaker will rate him more unfavorably than subjects administering neutral feedback. Eighty-two students enrolled in classes of public speaking courses were randomly assigned to one of three response-role conditions; Favorable, Neutral, or Unfavorable. In each class each member of the audience was given a set of cards to indicate his response-role. During the delivery of a speech given by a confederate, the audience was asked to rate the speakers. The results indicated that Amato and Ostermeier's findings lend them- selves to a response-role, rather than a speaking per- formance interpretation. Subjects administering positive feedback rated the speaker significantly higher than those in the neutral condition. Subjects administering negative feedback rated the speaker significantly lower than those in the neutral condition. It is apparent that the findings of these two experi- ments lend themselves to a third interpretation, one in line with self-perception theory. Audience members may have inferred their attitude toward the unknown speaker by observing their own overt behavior. Taken together, the studies cited above support the Bemian notion that attitudes can follow behavior. Finally, Bem asserts that individuals not only use the behavior of another individual as a guide for infer- ring his beliefs and attitudes, but also take into account the circumstances that appear to be responsible for his behavior. Accordingly, Bem's theory of self-perception also predicts that an individual not only uses his own 10 behavior as a guide for inferring his beliefs and attitudes, he also takes into account the circumstances that appear to be responsible for his behavior. Bem traces his analysis to Skinner's distinction between a mand and a tact: A descriptive statement, a verbal response that is under the discriminative control of some portion of the environment, is classified as a "tact" (Skinner, 1957). A speaker is trained to describe or "tact" his environment for the benefit of his listeners who pro- vide generalized social reinforcement in return. . . . Verbal responses that are under the control of specific reinforcing contingencies are called "mands." A speaker who emits a mand is asking for, requesting, or "manding" a particular reinforcer. . . . Mands are often dis- guised as tacts, as in "I believe you have the sports page" or as in the case of the television announcer who praises the product he is selling; his verbal be- havior is a mand for the salary he receives and may not at all be under the actual discriminative control that features the product he appears to be tacting. . . . It is clear, then, that in an attempt to infer a speaker's "true" beliefs and attitudes, the listener must often discriminate the mand-tact characteristics of the communication (1965, pp. 200-201). Thus, control over an individual's beliefs and attitudes exerted by his overt behavior is mitigated to the extent that cues are present implying that the behavior is de- ceitful or, more generally, is being emitted for immediate specific reinforcement. Bem has used this latter premise as a primary basis for his analysis of experiments conducted as tests of cognitive dissonance theory in which subjects encode counter-attitudinal messages. This premise, that rein- forcement (i.e., compensation for performing a behavior) mediates the subsequent amount of attitude change, is the focus of the next section of this paper. 11 An Alternative Explanation of Dissonance Phenomena Generally, investigators concerned with counter- attitudinal advocacy have concentrated on the effects of varying amounts of incentives used to induce a subject to encode a belief-discrepant communication. The foundation for such research can be found in Festinger's (1957) origi- nal formulation of the relationship between the amount of incentives and the magnitude of dissonance. The magnitude of the reward or punishment, that is, the attractiveness and desirability of the offered reward or the unpleasantness and the undesirability of the threatened punishment is an important determi- nant of the magnitude of dissonance which exists once compliance is exhibited. Too great a reward or punishment will result in only little dissonance (p. 91). Several studies have provided empirical support for Festinger's formulation (e.g., Festinger and Carlsmith, 1959; Cohen, 1962). These studies, as well as others, suggest that the persuasive efficacy of counter-attitudinal advocacy rests upon establishment of the minimal justifying conditions to induce belief-discrepant behavior. Not sur- prisingly, however, the dissonance position has come under attack. For certain aspects of reinforcement learning theories imply that belief-discrepant arguments are more likely to be accepted if the reward for encoding them is substantial. Moreover, studies by Scott (1957, 1959); Bostrom, Vlandis and Rosenbaum (1961); Janis and Gilmore (1965); and Elms and Janis (1965) indicate that after counter-attitudinal advocacy, more self-persuasion occurs 12 under conditions of high, rather than low justification. These researchers hold that if self-persuasion is the desired end, persons should be given positive incentives for encoding counter-attitudinal messages. Bem (1965, 1967a, 1967b, 1968, 1970a, 1970b) has joined the counter-attitudinal advocacy controversy. He assumes the dissonance results to be reliable, but offers a radically different mechanism for the effects. Bem pro- poses that the basic findings of both the "forced compli- ance" type of dissonance experiment (e.g., Festinger and Carlsmith, 1959) and the "free choice" type (e.g., Brehm, 1956) can be explained without recourse to the motivational interpretation preferred by dissonance theorists. Bem's basic argument can be illustrated with the paradigmatic case of the forced-compliance experiment in which indi- viduals write counter-attitudinal essays for varying amounts of compensation. The "classical" finding of such studies, as noted earlier, is that there is a negative relationship between justification and self-persuasion, i.e., the less the justification, the greater the self- persuasion. The self-perception theory approaches these results by considering the vieWpoint of an outside observer who sees an individual writing an essay for little or no com- pensation; he can rule out financial incentive as a moti- vating factor and infer something about the individual's attitudes. He can use an implicit self-selection rule and 13 ask: "What must this man's attitude be if he is willing to behave in this fashion in this situation?" Accordingly, he can conclude that the individual holds an attitude con- sistent with the view that is expressed in the essay. On the other hand, if an observer sees an individual writing an essay for a large compensation, he can infer little or nothing about the actual attitude of that individual be- cause such an incentive appears sufficient to evoke the behavior regardless of the individual's private views. The observer's best guess, then, is to suppose that the individual's attitude is similar to that which would be expressed by anybody who was selected at random and asked for his opinion--the attitude of a control subject, in other words. Self-perception theory asserts that subjects in dissonance experiments are themselves behaving just like these hypothetical observers. They survey their own be- havior of writing the essay and then ask themselves: "What must my attitude be if I am willing to behave in this fashion in this situation?" Accordingly, they produce the same pattern of results as the outside observers: low-compensation subjects infer that they must agree with the arguments in their essay, whereas high-compensation subjects discard their behavior as a relevant guide to their "actual'' attitudes and express the same attitudes as the control subjects. The same kind of reasoning pre- dicts the differential effects of other variables in l4 cognitive dissonance experiments (e.g., justification and freedom of choice manipulations). This analysis has been tested several times with an experimental methodology known as "interpersonal simu- lation" (Bem, 1965, 1967a, 1967b, 1968). In these studies, an observer-subject is actually given a description of one of the conditions of a dissonance experiment and asked to estimate the attitude of the subject whose behavior is either described or actually overheard. As the self- perception theory predicts, the attitude estimates of observer-subjects in such simulations do, in fact, repro- duce the original dissonance findings. Bem (1965), for example, conducted an interpersonal replication of Cohen's (1962) study. Each subject in Bem's simulation simply received a description of the situation faced by a single subject in the original Cohen study: He was told that in the spring of 1959, there was a "riot“ at Yale University in which the New Haven police incurred the wrath of the undergraduates." He was then told that an undergraduate had written an essay entitled "Why the New Haven Police Actions Were Justified," an essay which was to be clearly in favor of the police side of the riots. According to the description, “the decision to write such an essay or not was entirely up to the student, and he was told that he would be paid the sum of $.50 (1.00) if he would be willing to do so. The student who was asked agreed to do so, and wrote such an essay" (p. 203). Bem's observers 15 were asked to estimate the attitude toward the police of the subject described to them. These estimates in the control, $.50, and $1.00 conditions were strikingly similar to the data actually obtained by Cohen. In his 1967 paper, Bem reports a similar inter- personal replication of the Festinger and Carlsmith (1959) study. Again the results were quite similar to those re- ported by Festinger and Carlsmith. Similar results are also reported by Jones (1966) in a study in which subjects' attitudes and observers' judgments were compared directly in the same experiment. The theory of self-perception receives only indirect support from Bem's interpersonal simulations. To obtain more direct support, Bem (1965) conducted an experiment in which "truth" and "lie" signals were "raised from birth" in the laboratory so that their meanings would be unam- biguous. Bem's assumption was that the small and large amounts of money in the dissonance experiments act re- spectively as "truth" and "lie" signals to the subjects, informing them whether or not to believe their own essays. Bem presented the subjects with a looseleaf notebook con- taining cartoons which the subject had previously rated as neutral on a scale ranging from very unfunny to very funny. After subjects had been seated in an acoustically tiled recording room, they received the following instructions over an intercom: l6 . . . We are going to be examining an individual's ability to judge whether the speaker on the tape is telling the truth or not. To do this, some of the things you will say on the tape will be true state- ments; others will be untrue. . . . After I ask you a question, I will start the tape recorder and you should answer the question into the microphone in front of you. Whenever I turn on the tape recorder, one of the two colored light bulbs in the ceiling fixtures will also go on automatically. If the amber light goes on . . . , you should make up an untrue answer and speak it into the microphone as convinc- ingly and naturally as possible. . . . The lighting circuitry is set to select the two colored lights automatically and in random sequence (pp. 212-213). This first part of the procedure was to "raise in the laboratory" external stimuli which had, in the past been associated with truth-telling or lie-telling. The next part of the procedure was designed to have the subject make factual statements in the presence of these external stimuli. Subjects again looked at some of the same car- toons and announced their personal decision that it was funny. The truth (or lie) light was turned on "Just so you will know when I turn the tape recorder on and off. . . . " Once the tape recorder (and the truth or lie light) was turned on, the subject then stated his previous decision aloud into the tape recorder. Subjects were then asked to re-rate the cartoons. Seven of the eight subjects were persuaded to a greater extent by comments made in the presence of the truth light than by comments made in the presence of the lie light. Bem interprets these findings to mean that the "truth" light acted just like the small sums of money in the dissonance experiments, signaling to the individual that 17 his behavior was an indication of his true attitude; the "lie" light acted like the large sums of money, telling the individual that his behavior was irrelevant to his true attitude. It is obvious that Bem's alternative explanation of dissonance phenomena is a broad theoretical statement of some impact. However, if the past history of controver- sies like this one is any guide, it seems unlikely that a "crucial" experiment for discriminating between the two theories will ever be executed. At this juncture each theory appears capable of claiming some territory not claimed by the other, and one's choice of theory in areas of overlap is diminishing to a matter of loyalty or aesthetics. Thus, if one sets aside the question of ob- servers and whether the statement is a "mand" or a "tact," it is difficult to distinguish between Bem's interpre- tation and a dissonance interpretation of the forced compliance phenomena. Rationale and Hypotheses The research evidence summarized above leads to the conclusion that an individual's belief and attitude state- ments may be predicted and controlled by manipulating his overt behavior and the stimulus conditions under which it occurs. Such reasoning suggests an experiment in which subjects are asked to provide either positive feedback (i.e., cues of approval) or negative feedback (i.e., cues 18 of disapproval) to an unknown speaker. Self-perception theory, as well as the findings of Amato and Ostermeier (1967) and Combs and Miller (1968), predicts that subjects in such a situation will infer their attitudes toward the quality of the speech by observing their own overt behavior. Thus, the following hypothesis was tested: H : Persons administering positive feedback to a speaker will rate his speech more favorably than persons administering negative feedback. While not systematically investigated in non-counter- attitudinal research, self-perception theory also predicts that control over an individual's attitudes exerted by his overt behavior is mitigated to the extent that cues are present implying that the behavior is being emitted for immediate specific reinforcement. Thus, subjects who are compensated for administering either positive or negative feedback will discard their behavior as a relevant guide to their "actual" attitudes toward the quality of the speech and express the same attitudes as control subjects. Conversely, subjects who receive little or no compensation for administering the feedback have no justification for discarding their behavior as a relevant guide to their behavior. Based on such reasoning, the following hypothesis was tested: Hz: (a) Persons administering positive feedback to a speaker under conditions of low incentive Will rate his speech more favorably than persons administering positive feedback under conditions of high incentive; conversely, 19 (b) Persons administering negative feedback to a speaker under conditions of low incentive will rate his speech more unfavorably than persons administering negative feedback under conditions of high incentive. Of secondary interest in this investigation was the effect of administering positive or negative feedback under conditions of high or low incentive on subsequent credi- bility ratings of the speaker and attitude change toward the issue discussed in the speech. While not central, they were included in the belief that they might be affected by the relationships between behavior, its con- trolling circumstances, and subsequent inferences regarding attitudes. Based on the same reasoning that led to Hy- potheses 1 and 2, the following relationships were also predicted: H ° Persons administering positive feedback to a speaker will rate him more credible than persons administering negative feedback. (a) Persons administering positive feedback to a speaker under conditions of low incentive will rate the speaker more credible than persons administering positive feedback under conditions of high incentive; conversely, (b) Persons administering negative feedback to a speaker under conditions of low incentive will rate the speaker less credible than persons administering negative feedback under conditions of high incentive. H : Persons administering positive feedback to a speaker will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering negative feedback. H 6' 20 (a) Persons administering positive feedback to a speaker under conditions of low incentive will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering positive feedback under conditions of high incentive; conversely, (b) Persons administering negative feedback to a speaker under conditions of low incentive will demonstrate greater attitude change counter to the position taken in the speech than persons administering negative feedback under conditions of high incentive. CHAPTER II EXPERIMENT I* Overview On the basis of pretest results, an issue was chosen which was approximately normally distributed for Se. gs were randomly assigned to the experimental groups. They were asked to be "confederates" in a study allegedly designed to investigate the effects of audience feedback on communication behavior. In the experimental conditions, gs administered positive or negative feedback to an unknown speaker under conditions of high or low incentive. Pretest gs had judged the speech to be "average" and the speaker's credibility to be "moderate." gs then completed posttest questionnaires. In addition to containing attitude items dealing with the experimental issue, the posttest question- naire included items designed to measure reactions to the speech, the credibility of the speaker, incentive, and type of feedback administered. In order to determine the *I am grateful to the following colleagues for serving as Es in this investigation: Edward Bodaken, Blaine Goss, Gerhard Hanneman, Bonita Perry, and Edward Wotring. In addition, I would like to thank Donald Cushman who allowed his class to be used for the experi- ment. 21 22 effect of the speech itself, in the absence of administer- ing feedback or receiving an incentive for such behavior, a control condition was included in the design. SS in the control condition completed pretest instruments, listened to the speech, and then completed posttest instruments. Subjects Experimental §s were eighty-five students enrolled in an undergraduate communication course at Michigan State University. Control §s were sixty-five students enrolled in a similar course at the same institution. A university- wide student strike depleted this original subject pool to forty-six experimental §s and twenty control gs. In addition, attrition between the pretest and posttest reduced the number of experimental gs to forty-five for the analysis of the attitude change scores only. Procedures Pretest §s were administered a pretest questionnaire, ostensibly to solicit student opinion on possible issues to be used in a class project (see Appendix A). The instrument contained several issues believed to be salient and somewhat ego-involving for undergraduate students. The issue producing the most normal range of scores called for the use of early psychological testing to identify 23 children with criminal potential; thus, it was chosen as the experimental issue for the investigation. Independent Variables Two independent variables were manipulated in this investigation: 1. Incentive. This variable was dichotomized as High Incentive and Low Incentive. Se in the High Incentive conditions were told that in return for their help as "confederates" in this investigation they would not have to take a previously scheduled examination, that they would receive credit for the examination without taking it. In addition, they were told that they would be paid $2.50 for their assistance. .§$ in the Low Incentive conditions were simply asked to serve as "confederates." They were not offered compensation for their assistance. 2. Feedback. This variable was dichotomized as Positive Feedback and Negative Feedback. §S were instructed to assume the role of an individual audience member andpro- vide the speaker with nonverbal feedback according to an assigned response role. §$ in the Positive Feedback con- ditions were instructed to respond favorably to the speaker while gs in the Negative Feedback conditions were instructed to respond unfavorably. gs accomplished their feedback task by holding up 3" x 5" index cards every 30 seconds for about 10 seconds. §S who administered positive feedback to the 24 speaker displayed green cards while §$ who administered negative feedback displayed red cards. Inductions and Posttest Approximately three weeks after the pretest had been administered and the experimental issue had been chosen, gs were induced to perform the task of providing Positive or Negative Feedback to the speaker. E gave the following instructions to §S in the Low Incentive conditions: Several students are conducting a little pilot study to investigate the effects of feedback on communication behavior. What we do is select a number of speakers and then ask audience members to respond to them in a speci- fied manner. We tape record these speeches under various audience feedback conditions and then analyze them for things, such as nonfluencies, rate and verbal output. Professor has consented to allow us to solicit your services as audience members, that is, as our confederates in this little study. In this particular speaking situation we are asking you to assume the role of an individual audience member and provide the speaker with nonverbal feedback accord- ing to an assigned role. You have been given a GREEN (RED) card which will assist you in performing your response role. The GREEN (RED) card represents a favor- able (unfavorable) response. That is, each of you will provide the speaker with positive (negative) feedback during his speech. The speaker is aware that a GREEN (RED) card indicates that you approve (disapprove) of his communication behavior. The speech you will hear is six minutes long. Every 30 seconds you will hold up your card for about 10 seconds indicating that you approve (disapprove) of the speaker and his speech. In order to raise and lower your card at the appropriate time, just follow the lead of my assistant who will be standing behind the speaker attending to the tape recorder. Are there any questions concerning this procedure before the speaker arrives to present his speech? 25 gs in the High Incentive conditions were given the following instructions: The Department of Communication at Michigan State University has received a large grant from the federal government to investigate the effects of feedback on communication behavior. What we do is select a number of speakers and then ask audience members to respond to them in a specified manner. We tape record these speeches under various audience feedback conditions and then analyze them for things, such as nonfluencies, rate, and verbal output. Professor has volunteered your services as audience members, that is, as our confederates in this investigation. Your cooperation is essential to the success of our project. In fact, it is so important that Professor has authorized me to tell you that in return for your help you will not have to take the examination that he has scheduled for Friday, that you will receive credit for the exam without taking it. In addition, the provisions of our grant enable us to pay you $2.50 for assisting us in this study. gs in the High Incentive conditions then received exactly the same instructions regarding the administration of the feedback as Se in the Low Incentive conditions. It is important to note that while the preceding instructions occurred simultaneously for all experimental groups, the experimental speech itself did not. This is because the experimental design required that the same live speaker present approximately the same manuscript speech to each experimental group. The speech was approxi- mately six minutes long. Therefore, there was a time lag between the explanation of the experimental task and the actual presentation of the speech. There was no time lag in the Low Incentive:Positive Feedback condition. The time lag in the other three experimental groups ranged from 10 26 minutes in the Low Incentive:Negative Feedback condition to 30 minutes in the High Incentive:Negative Feedback con- dition. The speaker, a male confederate of E, was introduced in each condition as an undergraduate volunteer. He was unknown to the audience prior to the completion of his speech. The speech was prepared by E,in collaboration with the speaker and delivered from a typed manuscript (see Appendix B). During its delivery, the speaker maintained eye contact with the audience, since he had participated in several pretest sessions prior to the presentation and was familiar with the content of the speech. Pretest Es had judged the speech to be "average" and the speaker's credibility to be "moderate." Immediately following the speech Es completed post- test questionnaires. The instrument contained the same scales employed in the pretest to measure attitudes toward early psychological testing. In addition, it included items designed to measure reactions to the speech, the credibility of the speaker, and to assess the effiCacy of the feedback and incentive manipulations (see Appendix C). Immediately following the completion and collection of the posttest questionnaires, E explained the true nature of the study. In addition, in the High Incentive conditions E explained that it was true that in return for their help Es would receive credit for the examination without taking 27 it. E also explained that it would be impossible to pay them the $2.50 as stated in the task instructions. In the Low Incentive conditions E explained that in return for their help Es would also receive credit for the examination without taking it. At the end of the debriefing session the Es were thanked for their help and dismissed. E gave the following instructions to Es in the control condition: The Department of Communication at Michigan State University is attempting to construct a measuring in- strument that will validly rate several dimensions of a communication transaction. We would like for you to listen to a six minute speech and then rate it on some scales that we are developing. Are there any questions before I introduce the speaker? After listening to the speech, control Es completed a posttest questionnaire. The instrument was similar to the one administered to experimental Es except that it did not contain items to measure incentive and type of feed- back administered (see Appendix D). Dependent Variables Three dependent variables were measured in this investigation: 1. Speech Ratings. The posttest questionnaire con- tained speech ratings along four evaluative dimensions: Content and Analysis, Delivery, Language, and Overall Effectiveness. Each dimension was accompanied by a brief statement listing some of the criteria to be used in re- acting to it. A seven-interval, semantic differential-type, 28 rating scale appeared below each dimension. The scales used Superior, Excellent, Good, Average, Fair, Poor, and Very Poor as anchors. Prior research on commonly used speech rating scales had indicated that the Content and Analysis, Delivery, and Language dimensions were somewhat independent (Becker, 1962). The Overall Effectiveness dimension was included to obtain general audience reactions to the total speech. Each dimension was scored on a one low (least favorable), seven high (most favorable) basis. Therefore, the higher the speech rating, the more favorable the reaction. 2. Credibility. The posttest questionnaire con- tained credibility ratings along three evaluative dimen- sions: Qualification, Safety, and Dynamism. For each dimension, four, seven-interval, semantic differential- type rating scales were used. Scales employed for Quali- fication were bounded by the adjectives skilled-unskilled, qualified-unqualified, experienced-inexperienced, and informed-uninformed. Those employed for Safety were kind- cruel, honest-dishonest, friendly-unfriendly, and just- unjust. Scales employed for Dynamism were bounded by the adjectives bold-timid, energetic-tired, active-passive, and aggressive-meek. These scales were selected on the basis of factor analytic studies (Berlo, Lemert, and Mertz, 1970). Since credibility ratings were obtained by summing across the scales, a score of 28 indicated 29 highest credibility on each dimension, a score of four lowest on each dimension. 3. Attitude Change. The posttest questionnaire contained the same four semantic differential-type scales employed in the pretest to measure attitudes toward early psychological testing to identify children with criminal potential. The scales used were good-bad, valuable- worthless, pleasant-unpleasant, and fair-unfair (Osgood, Suci, and Tannenbaum, 1957). Since attitude scores were obtained by summing across the scales, a score of 28 indi- cated the most favorable attitude, a score of four the least favorable. Attitude change was defined as the difference between Es' attitude scores on the pretest and posttest measures. Manipulation Checks The effectiveness of the High Incentive-Low Incen- tive manipulations was determined by having Es respond to the following question: How adequate was the incentive for you to act as a confederate in this study? Es responded on a seven-interval, semantic differential-type scale which were bounded by the adjectives adequate-inadequate. Perceptions of the feedback manipulations were deter- mined by having Es respond to the following question: In your role as a confederate in this study, what type of feedback were you asked to administer to the speaker? Es responded on a scale consisting of three response categories: 30 (l) I was instructed to administer positive (favorable) feedback to the speaker; (2) I was instructed to administer negative (unfavorable) feedback to the speaker; and (3) I do not know what kind of feedback I was instructed to administer to the speaker. Results Manipulation Checks Incentive.-—In order to determine if subjects in each of the incentive conditions accurately perceived the level of the variable to which they were assigned (High Incentive or Low Incentive) and to discover whether or not these perceptions were significantly different, a two- factor analysis of variance was performed on the incentive perception ratings. This analysis (Table 1) yielded a significant incentive main effect. Subsequent Eftests demonstrated that the two levels of the incentive variable were perceived as significantly different (Table 2); thus, the manipulation of the incentive variable was successful. Feedback.--The effectiveness of the feedback manipu- lation was assessed by computing the percentage of indi- viduals within each experimental group who accurately perceived the experimental induction. These percentages are reported in Table 3. The figures in this table indi- cate that with the exception of the Low Incentive:Negative Feedback condition, 100 per cent of the subjects perceived 31 TABLE 1. Means, standard deviations, and analysis of variance of incentive ratings.* . . Standard Condition Mean n Deviation High Incentive:Positive Feedback 5.90 11 1.04 Low Incentive:Positive Feedback 3.58 12 1.88 High Incentive:Negative Feedback 6.45 11 .52 Low Incentive:Negative Feedback 3.16 12 1.26 Source of Variation SS df MS F Incentive 7.87 l 7.87 54.10** Feedback .00 1 .00 <1 n.s. Incentive x Feedback .23 1 .23 <1 n.s. Error 5.88 32. .14 Total 13.98 45 *Incentive perception scores ranged from 1 (Very Therefore, the higher the rating, the higher the incentive. Inadequate) to 7 (Very Adequate). **p < .05 32 TABLE 2. Estest comparisons of incentive ratings of experimental groups. Condition* High: + Low: + High: - Low: - Means 5.90 3.58 6.45 3.16 High: + 5.90 --—- 3.45** <1 5.39** Low: + 3.58 ---- 4.67** <1 High: - 6.45 ---- 7.65** Low: - 3.16 ---- *High: + = High Incentive:Positive Feedback; Low: + = Low Incentive:Positive Feedback; High: — = High Incentive: Negative Feedback; Low: - = Low Incentive:Negative Feedback. **p < .05 TABLE 3. Percentage and frequency of subjects accurately perceiving the feedback condition. Italics indicate intended inductions. Condition* Positive Negative Do Not Know High: + 100% (11) Low: + 100% (12) High: - 100% (11) Low: - 93% (ll) 7% (1) *High: + = High Incentive:Positive Feedback; Low: + = Low Incentive:Positive Feedback; High: - = High Incentive: Negative Feedback; Low: - = Low Incentive:Negative Feedback. 33 the type of feedback they were instructed to administer to the speaker as intended. In the Low Incentive:Negative Feedback condition 93 per cent of the subjects accurately perceived the experimental induction. Thus, the manipu- lation of the Feedback variable was successful. Test of Primarngypotheses: Speech Ratings Two preliminary analyses were performed on the speech rating scores along the four evaluative dimensions. First, mean ratings and standard deviations for experimental and control groups were computed (Table 4). Second, speech rating scores of the experimental and control groups were submitted to a simple analysis of variance (Table 5). This analysis demonstrated that the five groups differed signifi- cantly along the four speech rating dimensions. Within each dimension the specific differences between each experimental group and the control group were computed using Scheffe's test (McNemar, 1969). The results of these comparisons are reported in Table 6. They indicate that within each dimension the High Incentive:Positive Feedback condition, the Low Incentive:Positive Feedback condition, and the Low Incentive:Negative Feedback condition were not significantly different from the control group. Exami- nation of Table 6 reveals, however, that within each dimension, with the exception of Delivery, subjects in the High Incentive:Negative Feedback condition rated the 34 TABLE 4. Mean ratings and standard deviations for experi- mental and control groups within the four speech rating dimensions. Rating Dimension Condition* Mean Sgiigiign Content-Analysis High: + 4.90 .70 Low: + 4.75 .96 High: — 2.90 .99 Low: - 4.41 1.03 Control 4.85 1.03 Delivery High: + 4.63 1.12 Low: + 4.41 .99 High: - 3.20 1.13 Low: - 4.33 1.43 Control 4.50 1.19 Language High: + 4.54 1.03 Low: + 5.08 .28 High: - 3.50 .97 Low: - 4.08 1.16 Control 5.05 .88 Overall High: + 4.18 1.32 Effectiveness Low: + 3.91 1.16 High: - 2.70 1.05 Low: - 3.58 1.24 Control 4.25 .96 *High: + = High Incentive:Positive Feedback; Low: + Low Incentive:Positive Feedback; High: Negative Feedback; Low - = Low Incentive:Negative Feedback. High Incentive: 35 TABLE 5. Summary table of analysis of variance of speech ratings of experimental and control groups. Rating Source of Dimension Variation SS df MS F Content- Between 30.62 4 7.65 9.27* Analysis Within 49.52 E2 .82 Total 80.15 64 Delivery Between 14.28 4 3.57 2.52* Within 84.72 EE. 1.41 Total 99.01 64 Language Between 22.05 4 5.51 6.61* Within 50.01 E9_ .83 Total 72.06 64 Overall Between 18.46 4 4.61 3.59* Effective- Within 77.31 EE_ 1.28 ness Total 95.78 64 *p < .05 36 TABLE 6. Analysis of differences of speech ratings between each experimental group and control. Rating Dimension Condition* Control Means 4.85 Content-Analysis High: + 4.90 <1 n.s. Low: 4.75 <1 n.s. High: 2.90 5.56** Low: 4.41 <1 n.s. Means 4.50 Delivery High: 4.63 <1 n.s. Low: 4.41 <1 n.s. High: 3.20 <1 n.s. Low: 4.33 <1 n. . Means 5.05 Language High: 4.54 <1 n.s. Low: 5.08 aummmzum>aucmocH 30a I “30a “Monacoom m>aummmzum>aucmocH swam v mes .xomnemmm u I "swam “Momncmmm m>auamomum>aucmocH 3oq n + "30a axomncmom m>auamomum>aucmocH scam u + "swans IIII mm.m uBoa av III- o>.~ "seam av ssmvum IIII Hmum ."qu mmmc av ..me ~ av IIII ma 4 .a .m Im>aaomaam mm.m on.m Hm.m ma.v mane: Hamam>o av IIII om.m "Beam m>.m «*ma.m IIII mo.m "30a av v.m~.~ av III- vm.e "swam mo.e om.m mo.m em.e mammz mmmsmcma IIII mm.v "30a ssmm.H IIII om.m "coax av ««mm.m IIII He.v "304 av ..mk.~ av III- no.4 "seam mm.s om.m aa.e mm.e mamas sam>aamo IIII av.v "30a ssho.v IIII om.N "SOHE av ssmm.v IIII mn.e "30a av .«ma m av IIII om e "seam mamsamaa ae.e om.m mn.v om.v mcmmz Iucmucoo . . . . COHmeEaQ I .3oq I .cmam .3oa .nmam «c0auaccou mcaumm .mmdoam Hmucmfiaaomxm mo mosauma commmm came mo mcomanmmfioo ummuIm .m mqmda 40 (a) Persons administering positive feedback to a speaker under conditions of low incentive will rate his speech more favorably than persons administering positive feedback under conditions of high incentive; and conversely, (b) Persons administering negative feedback to a speaker under conditions of low incentive will rate his speech more unfavorably than persons administer- ing negative feedback under conditions of high incentive. In order to test this hypothesis, multiple comparison Eftests were performed on all possible two treatment combi- nations of the experimental groups along each speech rating dimension. The results of these comparisons indicate that Hypothesis 2 is not supported (Table 8). In fact, as noted earlier, subjects in the High Incentive:Negative Feedback condition rated the speech on the dimensions of Content and Analysis and Delivery significantly more unfavorably than subjects in the Low Incentive:Negative Feedback condition. It will be recalled that subjects in the High Incentive:Negative Feedback condition rated the speech on the dimension of Content and Analysis significantly more unfavorably than subjects in the control condition (Table 6). Thus, along the dimension of Content and Analysis, the results of Hypothesis 2(b) are in the opposite direction of that which was predicted. Test of Secondary Hypotheses: Credibiligy Two preliminary analyses were performed on the credibility rating scores along the three dimensions. First, mean ratings and standard deviations for 41 experimental and control groups were computed (Table 9). Second, credibility rating scores of the experimental and control groups were submitted to a simple analysis of vari- ance (Table 10). This analysis demonstrated that no sig- nificant differences existed in the means of the five groups on the three credibility dimensions. The third hypothesis of this study predicted a sig- nificant Feedback main effect along the three credibility dimensions; specifically: Persons administering positive feedback to a speaker will rate him more credible than persons administer- ing negative feedback. In order to test this hypothesis, a two-factor analysis of variance was performed on the data for the three rated dimensions of credibility. The results of this analysis yielded ambiguous support for Hypothesis 3 (Table 11). Of the three credibility dimensions, Safety was the only one to produce a significant Feedback main effect. Since, however, a previous analysis demonstrated that no significant differences existed in the means of the eXperimental and control groups (Table 10), it cannot 'be concluded that Hypothesis 3 is supported along the Safety dimension of credibility. The fourth hypothesis of this study predicted that the incentive variable would mediate the effect of adminis- tering feedback on subsequent credibility ratings; specifi- cally: 42 TABLE 9. Mean ratings and standard deviations for experi- mental and control groups within the three credibility dimensions. Credibilit . . Standard Dimensiony Condition* Mean Deviation Qualification High: + 18.72 3.82 Low: + 19.41 2.35 High: - 16.10 3.87 Low: - 18.91 2.23 Control 18.05 2.98 Safety High: + 19.09 3.30 Low: + 18.16 2.48 High: - 15.00 2.21 Low: - 17.16 2.40 Control 17.90 2.17 Dynamism High: + 17.81 3.84 Low: + 15.66 2.90 High: - 16.20 2.89 Low: — 18.08 2.31 Control 16.00 3.04 *High: + = High Incentive:Positive Feedback; Low: + = Low Incentive:Positive Feedback; High: - = High Incentive: Negative Feedback; Low: — = Low Incentive:Negative Feedback. TABLE 10. Summary table of analysis of variance of credibility ratings of experimental and control groups. R8tlng. soufce.0f ss df MS F Dimen51on Variation Qualifi- Between 71.15 4 17.78 1 n.s. cation Within 565.86 EE 9.43 Total 637.01 64 Safety Between 1159.47 4 289.86 1 n.s. Within 8846.76 EE 147.44 Total 10006.24 64 Dynamism Between 66.02 4 16.50 1 n.s. Within 546.08 E9 9.25 Total 612.10 64 43 TABLE 11. Summary table of two—factor analysis of variance of credibility ratings. Rating. Source of SS MS F DimenSion Variation Qualifi— Incentive (A) 3.07 1 3.07 <1 n.s. cation Feedback (B) 2.44 1 2.44 <1 n.s. A x B 1.13 1 1.13 <1 n.s. Error 27.36 E; .86 Total 33.90 44 Safety Incentive (A) .38 1 .38 <1 n.s. Feedback (B) 6.47 1 6.47 10.46* A x B 2.38 1 2.38 <1 n.s. Error 25.01 El .61 Total 34.24 44 Dynamism Incentive (A) .01 1 .01 <1 n.s. Feedback (B) .15 1 .15 <1 n.s. A x B 4.06 1 4.06 4.98* Error 33.21 E; .81 Total 37.43 44 *p < .05 44 (a) Persons administering positive feedback to a speaker under conditions of low incentive will rate ;the speaker more credible than persons administering positive feedback under conditions of high incentive; and conversely, (b) Persons administering negative feedback to a speaker under conditions of low incentive will rate the speaker less credible than persons administering negative feedback under conditions of high incentive. In order to test this hypothesis, multiple comparison E-tests were performed on all possible two-treatment combi- nations of the experimental groups along each credibility dimension. The results of these comparisons indicate that Hypothesis 4 is not supported (Table 12). In fact, subjects in the High Incentive:Negative Feedback condition rated the speaker on all three dimensions significantly less credible than subjects in the Low Incentive:Negative Feedback con- dition. It will be recalled, however, that no significant differences existed in the means of the experimental and control groups when they were submitted to a simple analy- sis of variance (Table 10). Thus, unlike Hypothesis 2(b), it cannot be concluded that the results of Hypothesis 4(b) are in the opposite direction of that which was predicted. Test of Secondary Hypotheses: Attitude Change After random assignment of subjects to experimental conditions, three preliminary analyses were performed on the attitude data. First, Bartlett's test for homogeneity of variance (Winer, 1962) was performed on the eXperimental and control groups' pretest scores. The results indicated 45 mo. v mes .xomnemme m>aummoznm>aucmocH Boa H I "30a “Romnpmwm m>aummmzum>apcmocH scam I "swam “xomncmmm m>auamomum>aucmocH Boa u + “30a “Romncmmm m>apamomum>aucmocH 50am n + "roams ..... mo.ma I "30a .Iam.m ..... om.ea I "swam *«hH.N Hv IIIII mm.mH + ”Boa av av av ..... am.ea + “seam mo.ma om.ma s©.ma aw.ka mammz sma5mcso ..... ma.sa I “33 ..mo.m IIIII oo.ma I "swam Hv «smm.m IIIII ma.ma + "BOA av «Ima.m av IIIII ao.ma + "Beam ma.ka oo.ma sa.ma mo.ma mamas samamm IIIII am.ma I "Boa .Imo.m IIIII oa.ea I ”swam av ..mm.m IIIII aa.ma + "30a av av av IIIII Ne.ma + ”Beam am.ma oa.ma aa.ma ma.ma mamas scaumoaaaamso I “Boa I "swam + .Boq .cmam ceauaccou ceamcmsao . . . . I . . saaaanaemao .mmsoam amucmEaammxm mo mmchma muaaanacmao cmmE mo mcomaammeoo ammuvm .Na mamas 46 no significant differences existed in the variances of the five groups (X2 = 1.21; p > .05). Second, a simple analysis of variance was performed on the pretest attitude scores of the experimental and control groups to ensure that the pretest scores were not significantly different. The results of this analysis indicated that the group means did not differ significantly (Table 13). Finally, attitude change scores of the experimental and control groups were submitted to a simple analysis of variance. The overall amount of change and analysis of variance results are found in Table 14. This analysis demonstrated that no significant differences existed be- tween the five groups on amount of attitude change. The fifth hypothesis of this study predicted a significant Feedback main effect on attitude change scores; specifically: Persons administering positive feedback to a speaker will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering negative feedback. In order to test this hypothesis, a two-factor analysis of variance was performed on the attitude change scores. The results of this analysis indicated that feed- back had no significant effect on subsequent attitude change. Thus, Hypothesis 5 was not supported (Table 15). The sixth hypothesis of this study predicted that the effect of administering feedback on subsequent attitude 47 TABLE 13. Means, standard deviations, and analysis of vari- ance of pretest attitude scores for experimental and control groups. . . Standard Condition Mean Deviation High Incentive:Positive Feedback 10.45 7.04 Low Incentive:Positive Feedback 15.75 8.46 High Incentive:Negative Feedback 12.20 6.64 Low Incentive:Negative Feedback 16.50 6.31 Control 15.70 7.78 soufce.°f ss df MS F Variation Between 328.16 4 82.04 <1 n.s. Within 3271.77 EE_ 54.52 Total 3599.93 64 48 TABLE 14. Attitude scores, amount of change, and analysis of variance of attitude change scores of experi- mental and control groups. Condition Pretest Posttest Change High: + 10.45 11.45 1.00 Low: + 15.75 15.75 .00 High: — 12.20 11.50 - .70 Low: — 16.50 12.66 -3.83 Control 15.70 15.35 - .35 soufce.0f ss df MS Variation Between 157.74 4 39.43 <1 n.s. Within 1916.31 60 31.93 Total 2074.06 64 49 TABLE 15. Summary table of two-factor analysis of variance of attitude change scores of experimental groups. Source of Variation SS df MS F Incentive (A) 4.26 1 4.26 <1 n.s. Feedback (B) 7.64 1 7.64 <1 n.s. A x B 1.13 1 1.13 <1 n.s. Error 122.13 E; 2.97 Total 135.20 44 TABLE 16. Eétest comparisons of mean attitude change scores of experimental groups. Condition High: + Low: + High: - Low: - Means 1.00 .00 -.70 -3.83 High: + 1.00 ---- <1 <1 <1 Low: + .00 <1 ---- <1 <1 High: - - .70 ---- <1 Low: - -3.83 ---- 50 change would be mitigated to the extent that cues are present implying that the behavior is being emitted for specific reinforcement; specifically: (a) Persons administering positive feedback to a speaker under conditions of low incentive will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering positive feedback under conditions of high incentive; and conversely, (b) Persons administering negative feedback under conditions of low incentive will demonstrate greater attitude change counter to the position taken in the speech than persons administering negative feedback under conditions of high incentive. In order to test this hypothesis, multiple comparison Estests were performed on all possible two treatment combi- nations of the experimental groups. The results of these comparisons indicated that Hypothesis 6 is not supported; no significant differences existed among the conditions (Table 16). Discussion Experiment I provides partial support for one of Bem's major theoretical premises, that on occasion an individual infers his own internal states by observing his own overt behavior. Inspection of the speech ratings, the primary data in this investigation, supplies strong evidence for this claim. Along each of the four rated dimensions of speaking effectiveness (Content and Analy- sis, Delivery, Language, and Overall Effectiveness) subjects administering positive feedback to the speaker rate his’speech significantly more favorably than persons 51 administering negative feedback. These results support the earlier findings of Amato and Ostermeier (1967) and Combs and Miller (1968). Inspection of the credibility ratings and attitude change scores, the secondary data in this investigation, provides ambiguous support for the hypothesis that people on occasion infer their attitudes from their behavior. Safety was the only dimension of credibility to signifi- cantly support the prediction that subjects administering positive feedback to a speaker will rate him more credible than persons administering negative feedback. The Qualifi- cation dimension of credibility approached significance; positive feedback means were higher than negative feedback means. Along the Dynamism dimension of credibility, how- ever, subjects who administered negative feedback to the speaker rated him more credible than subjects who adminis- tered positive feedback. Inspection of the attitude change score means revealed that persons administering positive feedback to a speaker demonstrated greater attitude change in the direction of the position taken in the speech than persons administering negative feedback. Thus, the pri- mary data in Experiment I, speech ratings, provide strong support for the notion that people on occasion infer their attitudes from their behavior while the secondary data, credibility ratings, and attitude change scores provide only partial support. 52 Bem's second major theoretical premise, that control over an individual's attitudes exerted by his overt behavior is mitigated to the extent that cues are present implying that the behavior is being emitted for immediate specific reinforcement, receives no support whatsoever from Experi- ment I. Inspection of the speech ratings, the primary data in this investigation, supplies no evidence for this claim. In fact, examination of the means along the four rated dimensions of speaking effectiveness revealed that a reinforcement effect occurred. For three of the four rated dimensions, subjects administering positive feedback to a speaker under conditions of high incentive rated the speech somewhat more favorably than persons administering positive feedback under conditions of low incentive. This reinforce- ment effect is even more marked in the negative feedback conditions. For all four dimensions, subjects administer- ing negative feedback to a speaker under conditions of high incentive rated his speech more unfavorably than subjects' administering negative feedback under conditions of low incentive. In addition, the High Incentive:Negative Feed- back condition shows the most powerful effect; i.e., ratings of subjects in that condition differed most markedly from control group ratings. Examination of the credibility ratings and attitude change scores, the secondary data in this investigation, also revealed a reinforcement effect rather than a self- perception one. Of the six critical mean comparisons made 53 along the three credibility dimensions, five support a reinforcement interpretation of the data. In contrast, of the two critical comparisons made with the means of the attitude change score data, one produced a self-perception effect (High Incentive:Positive Feedback versus Low In- centive:Positive Feedback) while the other produced a reinforcement effect (High Incentive:Negative Feedback versus Low Incentive:Negative Feedback). In summary, the data of the present study consistently shows that control over an individual's attitudes exerted by his overt behavior is reinforced to the extent that compen- sation is provided for the behavior. This finding, of course, is opposed to self-perception predictions. The results of Experiment I should be approached with caution. Several of the results--especia11y the ones per- taining to incentive as a mitigator of the extent to which one's attitudes are inferred from one's behavior--are in the Opposite direction of self-perception theory, a theory of potential importance. In addition, a university-wide student strike depleted the original subject pool to a level that mitigated against meaningful statistical analy- sis. The maximum number of subjects in each cell was twelve. Finally, it is of potential import that the experimental speech was of "average" quality across all experimental conditions. While pretest subjects had judged the speech to be "average," it is likely that a practice effect occurred. That is, despite a conscientious effort 54 to the contrary, it is reasonable to believe that the speech became better each time it was delivered. The control group data, which was obtained last, supports this conclusion to some extent. Moreover, the speech may not depart markedly enough from the behaviors. Such considerations led to a replication of Experiment I with two major exceptions: (1) an attempt was made to obtain a more sufficient sample size, and (2) the quality of the experimental speech was systematically manipulated as a control variable. The next chapter presents the proce- dures and results of this partial replication. CHAPTER I I I EXPERIMENT I I * Overview Experiment II was a replication of Experiment I with two major exceptions: (1) an attempt was made to obtain a more sufficient sample size, and (2) the quality of the experimental speech was systematically manipulated. Rationale and Hypotheses The hypotheses for Experiment II were generated from the same rationale employed for Experiment I; namely, that an individual's attitude statements may be predicted and controlled by manipulating his overt behavior and the stimulus conditions under which it occurs. Accordingly, Experiment II investigated the following primary hypotheses: H : Persons administering positive feedback to a 1 speaker will rate his speech more favorably than persons administering negative feedback. *I am grateful to the following colleagues for serv- ing as Es in this investigation: Beverly Clark, Blaine Goss, Gerhard Hanneman, Edward Wotring, and Frank Zink. Special thanks is due to John Kochevar for serving as the speaker. Finally, I would like to thank the following graduate students who allowed their classes to be used for either the pretest or the experiment: Richard Parker, Mantha Vlahos, James Nyenhuis, Stuart Surlin, Sandra Crispin, Robert Dixon, and Bonita Perry. 55 H2: (a) (b) 56 Given a high quality speech, (1) Persons administering positive feedback to a speaker under conditions of low in- centive will rate his speech more favorably than persons administering positive feedback under conditions of high incentive; con- versely, (2) Persons administering negative feedback to a speaker under conditions of low in- centive will rate his speech more unfavor- ably than persons administering negative feedback under conditions of high incentive. Given a low quality speech, (1) Persons administering positive feedback to a speaker under conditions of low incen- tive will rate his speech more favorably than persons administering positive feedback under conditions of high incentive; conversely, (2) Persons administering negative feedback to a speaker under conditions of low incen- tive will rate his speech more unfavorably than persons administering negative feedback under conditions of high incentive. In addition, Experiment II investigated the following secondary hypotheses: H3 Persons administering positive feedback to a speaker will rate him more credible than persons administering negative feedback. II: (a) (b) Given a high quality speech, (1) Persons administering positive feedback to a speaker under conditions of low incen- tive will rate the speaker more credible than persons administering positive feedback under conditions of high incentive; conversely, (2) Persons administering negative feedback to a speaker under conditions of low incen- tive will rate the speaker less credible than persons administering negative feedback under conditions of high incentive. Given a low quality speech, (1) Persons administering positive feedback to a speaker under conditions of low incen- tive will rate the speaker more credible than persons administering positive feedback under conditions of high incentive; conversely, 57 (2) Persons administering negative feedback to a speaker under conditions of low in- centive will rate the speaker less credible than persons administering negative feedback under conditions of high incentive. H : Persons administering positive feedback to a speaker will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering negative feedback. H6: (a) Given a high quality speech, (1) Persons administering positive feedback to a speaker under conditions of low in- centive will demonstrate greater attitude change in the direction of the position taken in the speech than persons administer- ing positive feedback under conditions of high incentive; conversely, (2) Persons administering negative feedback to a speaker under conditions of low in- centive will demonstrate greater attitude change counter to the position taken in the speech than persons administering negative feedback under conditions of high incentive. (b) Given a low quality speech, (1) Persons administering positive feedback to a speaker under conditions of low in- centive will demonstrate greater attitude change in the direction of the position taken in the speech than persons administer- ing positive feedback under conditions of high incentive; conversely, (2) Persons administering negative feedback to a speaker under conditions of low in- centive will demonstrate greater attitude change counter to the position taken in the speech than persons administering negative feedback under conditions of high incentive. Subjects Es were 192 students enrolled in undergraduate com- munication courses at Michigan State University. Attrition 58 between the pre- and posttest reduced this number to 119 Es for the analysis of the attitude change scores only. During the debriefing session an attempt was made to dis- cover those Es who had known the speaker prior to the speech or who had participated in Experiment I. On this basis five subjects were discarded from the analysis. Procedures Pretest Es were administered the same pretest employed in Experiment I (see Appendix A). While more skewed than the distribution obtained in Experiment I, the issue producing the most normal range of scores was again the proposition that called for early psychological testing to identify children with criminal potential. This fact, plus con- siderations of comparability across experiments, led to its re-use in Experiment II. Independent Variables Three independent variables were manipulated in this investigation: 1. Incentive. This variable was dichotomized as High Incentive and Low Incentive. Es in the High Incentive conditions were told that in return for their help as "con- federates" in this investigation they would be given $2.50. Unlike Experiment I, they were 22E told that they would receive credit for an examination without taking it. Es in 59 the Low Incentive conditions were simply asked to serve as "confederates." They were not offered compensation for their assistance. 2. Feedback. This variable was dichotomized as Positive Feedback and Negative Feedback. As in Experi- ment I, Es were instructed to assume the role of an indi- vidual audience member and provide the speaker with non- verbal feedback according to an assigned response role. Es in the Positive Feedback conditions were instructed to respond favorably to the speaker while Es in the Negative Feedback conditions were instructed to respond unfavorably. Es accomplished their feedback task by holding up 3" x 5" index cards every 30 seconds for about 10 seconds. Es who administered positive feedback to the speaker displayed green cards while Es who administered negative feedback displayed red cards. 3. Speech Quality. This variable was dichotomized as High Speech Quality and Low Speech Quality. It was included in this investigation to assess the possibility that self-perception effects are partially dependent on the existence of conflicting cues; e.g., the presentation of positive feedback when communication is of relatively low quality. Establishment of such a conflict situation is consistent with the type of environment in which self- perception effects have frequently been observed; i.e., one involving counter-attitudinal advocacy. Employing the "average" speech used in EXperiment I as a baseline (see 60 Appendix B), a speech manuscript was written by the in- vestigator in collaboration with the speaker to represent the High Quality Speech; this was then modified to obtain a speech manuscript representing the Low Quality Speech (see Appendix E). For the High Quality Speech, an effort was made to keep the organization lucid through the use of clear transitional statements and logical flow of ideas. For the Low Quality Speech, the sentences within each of the three main divisions of the speech (introduction, body, and conclusion) were randomized. This procedure was com- parable to one employed by Thompson (1960). The High Quality Speech was written in a conversational style, for example, personal pronouns were used. The Low Quality Speech was written in an impersonal manner, a style similar to the one utilized in this report. Finally, the High Quality Speech was delivered with no nonfluencies or other delivery flaws. The Low Quality Speech was delivered in a similar manner except that a nonfluency was introduced approximately every fourteenth word. This procedure was comparable to one employed by Miller and Hewgill (1965). Both vocalized pauses and repetitions were included in alternating order. The High Quality and Low Quality Speeches were pre- tested in order to determine if they were perceived as intended and to discover whether or not these perceptions were significantly different. The results of this pretest 61 are reported in Table 17. While the High Quality Speech did not receive high or favorable mean ratings in an absolute sense (i.e., "superior" or "excellent"), given Es' slightly unfavorable attitude toward the speech topic, it was decided that a mean rating of "good" was the most realistic rating that could be achieved. Similarly, the Low Quality Speech did not receive low or unfavorable mean ratings in an absolute sense (i.e., "very poor"). Never- theless, it was decided that a mean rating of "poor" was adequate for the purposes of this investigation. Thus, the results of the pretest demonstrated that the two versions of the experimental speech were perceived as intended in a relative sense, and these perceptions were significantly different; thus, these two speeches were used in the experiment. Inductions and Posttest Approximately two weeks after the pretest had been administered and the experimental issue had been chosen, Es were induced to perform the task of providing Positive or Negative Feedback to the speaker. E gave the following instructions to Es in the Low Incentive conditions: Several students are conducting a little pilot study to investigate the effects of feedback of communication behavior. What we do is select a number of speakers and then ask audience members to respond to them in a specified manner. We tape record these speeches under various audience feedback conditions and then analyze them for things such as nonfluencies, rate, and verbal output. 62 may .maommnmna mo. v m«« .coauomma may magmao>mw wHOE can .mGHDMH may Hmcmac .AHOHHmmsmV b on Aaoom mam>v a Scam ommcma mmaoom mcauma zommmma mm. ma aa.~ sommmm wuaamso 30a mmmcm>auowmmm .Iem.k ma.a ma mo.e nommmm spaamso seam aamum>o o~.a ma es.~ aommmm saaamso 30a .Ima.s as. ea ma.m nommmm saaamso Beam momsmqma oo.a ma mw.a nommmm saaamso 30a ..mm.aa mm. ea ma.m nommmm suaamso seam snm>aamo we. ma mm.m sommmm saaamso 30a .«mm.e no.a ma am.m sommmm saaamso seam mamsamanucmucoo coanma>wo u camccmum a «new: commmm mo muaamso coamcman mcaumm .msowmcmeo mcfiumn nommmm udow ecu caspHB zoflwmm muaamsw 30H may msmHm> sommmm auaamnv amen ecu mo mQOmHHmmEoo ammulu can .mcoauma>mc oumoqmum .mcmma ummumHm .na mqmds 63 Your instructor has consented to allow us to solicit your services as audience members, that is, as our confederates in this little study. In this particular speaking situation we are asking you to assume the role of an individual audience member and provide the speaker with nonverbal feedback accord- ing to an assigned role. You have been given a GREEN (EEE) card which will assist you in performing your response role. The GREEN (RED) card represents a favorable (unfavorable) response. That is, each of you will provide the speaker with positive (negative) feedback during his speech. The speaker is aware that a GREEN (RED) card indicates that you approve (Eig- approve) of his communication behavior. The speech you will hear is six minutes long. Every 30 seconds you will hold up your card for about 10 seconds indicating that you approve (disapprove) of the speaker and his speech. In order to raise and lower your card at the appropriate time, just follow the lead of my assistant who will be standing behind the speaker attending to the tape recorder. Are there any questions concerning this procedure before the speaker arrives to present his speech? Es in the High Incentive conditions were given the follow- ing instructions: The Department of Communication at Michigan State University has received a large grant from the federal government to investigate the effects of feedback on communication behavior. What we do is select a number of speakers and then ask audience members to respond to them in a specified manner. We tape record these speeches under various audience feedback conditions and then analyze them for things, such as nonfluencies, rate and verbal output. ' Your instructor has volunteered your services as audience members, that is, as our confederates in this investigation. Your cooPeration is essential to the success of our project. In fact, it is so important that the provisions of our grant enable us to pay you $2.50 for assisting us in this study. That is, at- the end of the project this morning (afternoon) we will pay you $2.50 for your help. 64 Es in the High Incentive conditions then received exactly the same instructions regarding the administration of the feedback as Es in the Low Incentive conditions. In order to maximize and equalize the number of Es in the eight experimental conditions, as well as achieve partial randomization, classes that met in the same class- room building at the same time were combined. Thus, classes A, B, and C met at 8:00 a.m. in classroom building Y. The Es in these three classes were randomly assigned to four conditions. Similarly, classes D and E met at 9:10 a.m. in classroom building Y. The Es in these two classes were randomly assigned to two conditions. Likewise, classes F and G met at 10:20 a.m. in classroom building Y. The Es in these two classes were combined for one condition. Finally, class H met at 3:00 p.m. in classroom building Y. The Es in this class constituted one condition. Randomi- zation was achieved by distributing tickets to Es as they arrived for their regularly scheduled class. It is important to note that while the task in- structions occurred-simultaneously for all experimental groups within a particular time period, the experimental speech did not. This is because the experimental design required that the same live speaker present the appropriate manuscript speech (High Quality Speech and Low Quality Speech) to each experimental group. The two versions of the speech were approximately six minutes long. Therefore, there was a time lag between the explanation of the 65 experimental task and the actual presentation of the speech in four of the eight conditions. In the 8:00 a.m. time period there was no time lag in the High Quality Speech: Low Incentive:Positive Feedback condition. The time lag in the other three experimental conditions ranged from 10 minutes in the High Quality Speech: Low Incentive:Negative Feedback condition to 30 minutes in the High Quality Speech: High Incentive:Negative Feedback condition. In the 9:10 a.m. time period there was no time lag in the Low Quality Speech: Low Incentive:Positive Feedback condition. The time lag in the second condition administered during that time period, the Low Quality Speech: Low Incentive:Negative Feedback treatment, was 10 minutes. Since there was only one condition adminis- tered during the 10:20 a.m. and the 3:00 p.m. time periods, there was no time lag. The speaker, a male confederate of E, was introduced in each condition as one of a number of volunteers for this project. He was unknown to the audience prior to the com- pletion of his speech. Immediately following the speech Es completed post- test questionnaires. The instrument contained the same scales employed in the pretest to measure attitudes toward early psychological testing. In addition, it included items designed to measure reactions to the speech, the credibility of the speaker, and to assess the efficacy of 66 the feedback and incentive manipulations. It was the same posttest employed in Experiment I (see Appendix C). Immediately following the completion and collection of the posttest questionnaires, E explained the true nature of the study. In addition, in the High Incentive conditions E explained that it would be impossible to pay Es the $2.50 as stated in the task instructions. At the end of the de- briefing session Es were thanked for their help and dis- missed. Dependent Variables Like Experiment I, three dependent variables were measured in this investigation: (1) speech ratings; (2) credibility ratings; and (3) attitude change. These vari- ables were operationalized in exactly the same way as they were in Experiment I. Manipulation Checks The effectiveness of the High Incentive-Low Incentive manipulations and the Positive Feedback-Negative Feedback manipulations were measured in exactly the same way as they were in EXperiment I. Results Manipulation Checks Incentive.--In order to determine if subjects in each of the incentive conditions accurately perceived the 67 level of the variable to which they were assigned (High Incentive or Low Incentive) and to discover whether or not these perceptions were significantly different, a three- factor analysis of variance was performed on the incentive perception ratings. This analysis (Table 18) yielded a significant incentive main effect. Subsequent Eftests demonstrated that with the exception of the Low Quality Speech: High Incentive:Negative Feedback condition and the Low Quality Speech: Low Incentive:Positive Feedback condition, the two levels of the incentive variable were perceived as significantly different (Table 19); thus, the manipulation of the incentive variable was only partially successful. Feedback.--The effectiveness of the feedback manipu- lation was assessed by computing the percentage of indi- viduals within each experimental group who accurately perceived the experimental induction. These percentages are reported in Table 20. The figures in this table indi- cate that with the exception of the Low Quality Speech: Low Incentive:Negative Feedback condition, 100 per cent of the subjects perceived the type of feedback they were instructed to administer to the speaker as intended. In the Low Quality Speech: Low Incentive:Negative Feedback condition, 92 per cent of the subjects accurately perceived the experimental induction. Thus, the manipulation of the Feedback variable was successful. 68 TABLE 18. Means, standard deviations, and analysis of vari- ance of incentive ratings. . . Standard * ** Condition Mean n Deviation H-Q, H-I, + 5.26 23 1.67 H-Q' H-I’ - 5.18 16 lo 54 H-Q, L-I' + 3063 19 1075 H-Q, L-I, - 3.05 20 1.49 L-Q, H-I, + 5.44 49 1.44 L-Q, H-I, - 4.58 12 2.17 10.0, 10.1, + 4008 23 1055 L-Q, L—I, - 3.16 25 1.99 soufce.°f ss df MS F Variation Quality of Speech (A) .05 l .05 <1 n.s. Incentive (B) 15.13 1 15.13 5.20*** Feedback (C) 108.45 1 108.45 37.30*** AB 3.27 1 3.27 <1 n.s. AC 2.43 1 2.43 <1 n.s. BC .81 l .81 <1 n.s. ABC .50 1 .50 <1 n.s. Error 689.91 186 Total 820.98 193 *H-Q = High Quality Speech; L-Q = Low Quality Speech; H-I = High Incentive; L-I = Low Incentive; + = Positive Feedback; - = Negative Feedback. **Incentive perception scores ranged from 1 (Very Inadequate) to 7 (Very Adequate). Therefore, the higher the rating, the higher the incentive. ***p < .05 69 mo. v a. oo.m I.aIa.0Ia mo.w +.aIa.0Ia av av av av mm.v I.aIm.oIa 4mm.m «mm.m «oa.m «om.v vv.m +.HI:.OIa mo.m I.aIa.0Ir mo.m +.aIa.OIm 8mm.m av «mo.v sno.m ma.m I.aIr.0Ir ama.v «mv.~ cv¢.v soo.m o~.m +.aIr.oI: oo.m mo.v mm.q vv.m mo.m mm.m ma.m m~.m memo: I.aIa.0Ia +.aIa.0Ia I.aI:.0Ia +.aIm.oIa I.aIa.0Im +.aIa.0Im I.aIm.0Im +.aIm.oIm coauaecoo .mmsoum amucmsaummxm mo mmcaumu m>aucmoca came mo mnemaummeoo ammuLw .ma wands 70 TABLE 20. Percentage and frequency of subjects accurately perceiving the feedback condition. Italics indicate intended inductions. Condition Positive Negative Do Not Know H-Q, H-I, + 100% (23) H-Q, H-I, - 100% (16) H-Q, L-I, + 100% (19) H-Q, L-I, - 100% (20) L-Q, H-I, + 100% (49) L-Q, H-I, - 100% (12) L-Q, L-I, + 100% (23) L-Q, L-I, - 92% (23) 8% (2) 71 Test of Primary Hypotheses: Speech Ratings Mean ratings and standard deviations for each experimental group along the four evaluative dimensions of speaking effectiveness were computed (Table 21). The first hypothesis of this study predicted a significant feedback main effect along the four evaluative speech rating dimensions; specifically: Persons administering positive feedback to a speaker will rate his speech more favorably than persons administering negative feedback. In order to test this hypothesis, a three-factor analysis of variance was performed on the data for the four rated dimensions of speaking effectiveness. The re- sults of this analysis indicate that Hypothesis 1 is not supported (Table 22). The analysis did indicate, however, a significant Quality of Speech by Feedback interaction along the Delivery and Language dimensions of speaking effectiveness. Examination of the treatment means along the Delivery dimension revealed the following nonsymmetri- cal interaction pattern: persons administering positive feedback to a high quality speech rated the speech less favorably than persons administering negative feedback; BEE persons administering positive feedback to a low quality speech did not rate the speech significantly different than persons administering negative feedback. An opposite nonsymmetrical pattern produced-the significant Quality of Speech by Feedback interaction along the TABLE 21. 72 Mean ratings and standard deviations for experi- mental groups within the four speech rating dimensions. . . . . . Standard Rating DimenSion Condition Mean n Deviation Content-Analysis H-Q, H-I, + 5.21 23 1.06 H-Q, H-I, - 4.93 16 .82 H-Q, L-I, + 4.63 19 1.22 H-Q, L—I, - 4.50 20 1.07 L-Q, H-I, + 4.26 49 1.27 L-Q, H-I, — 4.41 12 1.32 L-Q, L-I, + 4.26 23 1.42 L—Q, L-I, - 4.20 25 .97 Delivery H-Q, H-I, + 5.30 23 .80 H-Q, H-I, - 5.56 16 .99 H-Q, L-I, + 4.89 19 1.41 H-Q, L-I, - 4.60 20 .96 L-Q, H-I, + 3.10 49 1.14 L-Q, H-I, - 3.33 12 1.17 L-Q, L-I, + 3.56 23 1.34 L-Q, L-I, - 3.80 25 1.23 Language H-Q, H-I, + 5.21 23 .83 H-Q, H-I, - 5.31 16 1.15 H-Q, L-I, + 5.05 19 .99 H-Q, L-I, - 4.60 20 .73 L-Q, H-I, + 3.98 49 .91 L-Q, H-I, - 3.91 12 1.18 L—Q, L-I, + 4.04 23 1.39 L-Q, L—I, - 4.36 25 1.05 Overall H—Q, H-I, + 4.65 23 1.16 Effectiveness H-Q, H-I, - 4.68 16 .84 H-Q, L-I, + 4.21 19 1.50 H-Q, L-I, - 4.05 20 1.11 L-Q, H-I, + 3.55 49 1.32 L-Q, H-I, - 3.33 12 1.10 L—Q, L-I, + 3.65 23 1.27 L-Q, L-I, - 3.52 25 1.38 73 TABLE 22. Summary table of three-factor analysis of variance of speech ratings. Rating Source of Variation SS df MS F Dimension Content- Quality of Speech (A) 11.60 1 11.60 8.01* Analysis Incentive (B) 3.91 l 3.91 <1 n.s. Feedback (C) .26 l .26 <1 n.s. AB .63 1 .63 <1 n.s. AC 1.62 l 1.62 <1 n.s. BC .01 1 .01 <1 n.s. ABC .37 l .37 <1 n.s. Error 280.74 186 Total 299.14 193 Delivery Quality of Speech (A) 108.75 1 108.75 78.42* Incentive (B) .49 1 .49 <1 n.s. Feedback (C) .46 l .46 <1 n.s. AB .63 l .63 <1 n.s. AC 13.38 1 13.38 9.65* BC .76 l .76 <1 n.s. ABC .78 l .78 <1 n.s. Error 397.45 186 Total 522.73 193 Language Quality of Speech (A) 30.08 1 30.08 34.48* Incentive (B) .34 1 .34 <1 n.s. Feedback (C) .02 1 .02 <1 n.s. AB .94 l .94 <1 n.s. AC 4.84 l 4.84 4.38* BC .07 l .07 <1 n.s. ABC 2.17 l 2.17 <1 n.s. Error 197.66 186 Total 236.13 193 Overall Quality of Speech (A) 31.72 1 31.72 19.11* Effective- Incentive (B) 1.58 l 1.58 <1 n.s. ness Feedback (C) .57 1 .57 <1 n.s. AB .12 l .12 auomwwm Nm.m mm.m mm.m mm.m mo.v aN.v mo.v mo.v meme: aauao>o av av av av av .mm.~ .nm.~ mm.v I.aIa.oIa av av av .nm.~ .am.m .nm.m vo.v +.aIa.oIa av av .mn.~ .ao.m .mo.m am.m I.aIr.oIa av cma.v «vw.v cov.m mm.m +.HI$.OIA av av .ne.~ oo.v I.aIa.oI: av av mo.m +.aIa.oI: av am.m I.HIm.OIm a~.m +.HIm.OI: mm.v vo.v am.m mm.m oo.e mo.m am.m a~.m meme: momsmcma av av av av «mm.~ .mv.v smm.v om.m I.aIa.oIa av av .om.~ .vo.m av .a~.m om.m +.aIa.oIa av «ma.m vmw.~ .am.m .mo.m mm.m I.aIr.oIa .no.m vmm.m «om.h .ma.m oa.m +.aIm.0Ia av smm.m amm.~ oo.v I.aIa.oI: av av mm.v +.aIa.oI= av om.m I.aIm.OIr om.m +.aIz.oIr om.m om.m mm.m oa.m ow.v mm.v om.m om.m meme: >um>aamo av av av av av av smm.m om.v I.HIa.oIa av av av av av vam.~ mm.v +.aIa.oIa av av av av av.v I.aIr.oIa av av soo.m om.v +.aI:.OIa av av av om.v I.aIa.oIm av av mm.v +.HIa.oI: av mm.¢ I.aIm.oI= AN m + HIE OI: mHm%HMC< o~.v o~.¢ av.¢ mm.v om.v mo.v mm.v a~.m meme: Iucmucoo I.aIa.oIa +.aIa.0Ia I.aI:.oIa +.aI=.oIa I.aIa.oIz +.aIa.oIm I.aI=.oI: +.aI=.oI: acauaecoo acawmwwwm .mmnoam amucmEaummxm mo monauma Loemmm name no mCOmauomEoo ummuIm .mN mamda 77 TABLE 24. Mean ratings and standard deviations for experi- mental groups within the three credibility dimensions. Credibilit . . Standard Dimensiony Condition Mean n Deviation Qualification H—Q,H-I,+ 20.87 23 2.38 H-Q,H-I,- 20.68 16 2.61 H-Q,L-I,+ 19.68 19 4.06 H-Q,L—I,- 18.35 20 3.16 L-Q,H—I,+ 16.34 49 3.27 L-Q,H-I,- 15.66 12 3.27 L-Q,L-I,+ 15.52 23 4.07 L-Q,L-I,- 16.48 25 3.25 Safety H-Q,H-I,+ 18.30 23 2.28 H—Q,H—I,- 17.75 16 2.17 H-Q,L-I,+ 18.15 19 2.58 H-Q,L-I,— 17.80 20 2.78 L-Q,H-I,+ 18.08 49 3.40 L—Q,H-I,- 17.66 12 3.25 L—Q,L-I,+ 17.82 23 2.67 L—Q,L—I,- 17.48 25 2.58 Dynamism H-Q,H-I,+ 17.08 23 3.92 H-Q,H—I,- 16.43 16 3.02 H—Q,L-I,+ 16.15 19 3.55 H—Q,L—I,- 15.60 20 2.72 L—Q,H-I,+ 14.32 49 2.41 L-Q,H—I,- 13.00 12 2.51 L-Q,L-I,+ 14.69 23 2.78 L-Q,L-I,- 14.92 25 3.34 78 The third hypothesis of this study predicted a significant feedback main effect along the three credi- bility dimensions; specifically: Persons administering positive feedback to a speaker will rate him more credible than persons administer— ing negative feedback. In order to test this hypothesis, a three-factor analysis of variance was performed on the data for the three credibility dimensions. The results of this analysis indicated that feedback had no significant effect on subse- quent credibility ratings (Table 25). Thus, Hypothesis 3 was not supported. The analysis did indicate, however, a significant Quality of Speech by Feedback interaction along the Dynamism dimension. Examination of the treatment means revealed the following nonsymmetrical interaction pattern: persons administering positive feedback to a high quality speech rated the speaker significantly more credible than persons administering negative feedback; BEE! persons administering positive feedback to a low quality speech did not rate the speaker significantly different than persons administering negative feedback. Thus, for those peOple administering feedback to a high quality Speech along the dimension of Dynamism, the results confirm Hypothesis 1. The analysis also yielded a significant Quality of Speech main effect along the dimensions of Safety and Dynamism. Persons exposed to a high quality speech rated the speaker more credible than persons exposed to a low quality speech. 79 TABLE 25. Summary table of three-factor analysis of variance of credibility ratings. Rating. Source of Variation SS df MS F DimenSion Qualification Quality of Speech (A) 2.31 1 2.31 l n.s. Incentive (B) .73 l .73 l n.s. Feedback (C) 7.07 l 7.07 1 n.s. AB .05 1 .05 1 n.s. AC .30 1 .30 1 n.s. BC .17 l .17 1 n.s. ABC .06 1 .06 1 n.s. Error 1455.79 EEE Total 1467.13 193 Safety Quality of Speech A 612.91 1 612.91 53.35* Incentive (B) 31.56 1 31.56 1 n.s. Feedback (C) 3.87 l 3.87 1 n.s. AB 8.13 1 8.13 1 n.s. AC 31.14 1 31.14 1 n.s. BC .59 1 .59 l n.s. ABC 19.67 1 19.67 1 n.s. Error 2799.23 186 Total 3507.19 193 Dynamism Quality of Speech (A) 175.73 1 175.73 18.39* Incentive (B) .69 l .69 l n.s. Feedback (C) 13.47 1 13.47 1 n.s. AB .02 1 .02 1 n.s. AC 41.55 1 41.55 4.35* BC 6.81 1 6.81 1 n.s. ABC 5.38 l 5.38 l n.s. Error 1940.86 186 Total 2184.53 193 80 The fourth hypothesis of this study predicted that the incentive variable would mediate the effect of administering feedback on subsequent credibility ratings; specifically: (a) (b) Given a high quality speech, (1) Persons administering positive feedback to a speaker under conditions of low incentive will rate the speaker more credible than persons administering positive feedback under conditions of high incentive; conversely, (2) Persons administering negative feedback to a speaker under conditions of low incentive will rate the speaker less credible than persons administering negative feedback under conditions of high incentive. ~ Given a low quality speech, (1) Persons administering positive feedback to a speaker under conditions of low incentive will rate the speaker more credible than persons administering positive feedback under conditions of high incentive; conversely, (2) Persons administering negative feedback to a speaker under conditions of low incentive will rate the speaker less credible than persons administering negative feedback under conditions of high incentive. In order to test this hypothesis, multiple comparison Estests were performed on all possible two treatment combi- nations of the experimental groups along each credibility dimension. The results of these comparisons indicated that Hypothesis 4 is not supported (Table 26). Test of Secondary Hypotheses: Attitude Change Two preliminary analyses were performed on the attitude data. First, Bartlett's test for homogeneity of 8]. mo.vnm¢ av av av av av av av ~m.va I.aIa.oIa av av av av av av aw.va +.HIA.OIA av aoo.m «mm.m anh.~ «ha.m oo.ma I.HIm.OIa av cmm.~ smh.m «ao.m Nm.va +~HIE~OIA av av av oo.ma I.aIa.OI= av av ma.oa +.ala.0Im av m¢.ma I.HIm.OI: mo.na +.aI=.oI: ~m.va mo.ea oo.ma ~m.ea om.ma ma.oa me.ma mo.ea mama: Emasmcso av av av av av av av mv.na I.aIa.oIa av av av av av av mm.ha +.HIa.OIa av av av av av mm.na I.HIm.OIa av av av av mo.wa +.HI=.OIA av av av om.na I.aIa.oIm av av ma.ma +.HIa.OI: av mh.ha I.HIm.OIm om.ma +.aI:.oI: me.sa am.ea em.ka mo.ma om.na ma.ma mA.Aa oa.ma mama: sumaam av av av av Nb.~ mo.v «mm.v mv.ma I.HIA.OIA av av amv.~ sa~.m amm.v «mm.m mm.ma +.HIa.OIa av av amn.m cvm.¢ sNN.m mw.ma I.HI:.OIa av sov.m cvs.v 4mm.m vm.wa +.HI=.OIa av av som.m mm.ma I.HIa.OIm av av mm.ma +.HIa.OI: av mo.o~ I.HI:.OI= hm.om +.HIm.OIm mo.ma ~m.ma em.ma em.ma mm.ma me.ma mm.om am.o~ mane: coaumUaaaamso Ialel slat lelal ale! lelsl .I.I III I I COflmCUEHQ a a o a + a a o a a m o a + a m o a a a o m + a a o m .a =.o : +.a =.o : cOauancoo auaaanaumuo .mm50nm amucwEanmmxe uo mocaumu huaaanacmuo came mo mnemaummeou umeuIm .mm mam .05). Second, a simple analysis of variance was performed on the pretest attitude scores to ensure that the pretest scores were not significantly different. The results of this analysis indicated that the group means did not differ significantly (Table 27). Mean pretest, posttest, and attitude change scores for the experimental groups are found in Table 28. The fifth hypothesis of this study predicted a significant Feedback main effect on attitude change scores; specifically: Persons administering positive feedback to a speaker will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering negative feedback. In order to test this hypothesis, a three-factor analysis of variance was performed on the attitude change scores. The results of this analysis indicated that feed- back had no significant effect on subsequent attitude change. Thus, Hypothesis 5 was not supported (Table 29). The sixth hypothesis of this study predicted that the effect of administering feedback on subsequent atti- tude change would be mitigated to the extent that cues are present implying that the behavior is being emitted for specific reinforcement; specifically: 83 TABLE 27. Means, standard deviations, and analysis of variance of pretest attitude scores for experimental groups. . . Standard Condition Mean n Deviation H-Q' H-I, 17.40 15 5.02 H-Q' H-I’ 13.33 12 7037 n.0, L-I' 13.50 18 5.76 H-Q' L’I’ 10.41 12 7.70 L—Q, H-I, 11.89 19 7.66 L-Q' H-I’ 10.58 12 6.82 L-Q' L-I’ 15.17 17 7.40 L-Q’ L‘I' 13007 13 6.73 Source of Variation Between 540.86 7 77.26 <1 n.s. Within 5044.02 109 46.27 Total 5584.88 116 TABLE 28 . 84 Mean pretest, posttest, and attitude change scores for experimental groups. Condition n Pretest Posttest Change H-Q, H-I, 15 17.40 18.53 1.13 H-Q, H-I, 12 13.33 13.00 -0.33 H-Q, L-I, 18 13.50 13.88 0.38 H-Q, L-I, 12 10.41 11.41 1.00 L-Q, H-I, 19 11.89 13.57 1.68 L—Q, H-I, 12 10.58 11.41 0.83 L-Q, L—I, 17 15.17 13.41 -l.76 L-Q, L-I, 14 13.50 11.85 -l.64 TABLE 29. 85 Summary table of three—factor analysis of vari— ance of attitude change scores of experimental groups. Source of Variation SS df MS Quality of Speech (A) 12.17 1 12.17 n.s. Incentive (B) 26.38 1 26.38 n.s. Feedback (C) .09 l .09 n.s. AB 44.35 1 44.35 n.s. AC 5.12 1 5.12 n.s. BC 4.41 1 4.41 n.s. ABC 2.61 1 2.61 Error 3830.64 186 Total 3925.79 193 86 (a) Given a high quality speech, (1) Persons administering positive feedback to a speaker under conditions of low incentive will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering positive feedback under conditions of high incentive; conversely, (2) Persons administering negative feedback to a speaker under conditions of low incentive will demonstrate greater attitude change counter to the position taken in the speech than persons administering negative feedback under conditions of high incentive. (b) Given a low quality speech, (1) Persons administering positive feedback to a speaker under conditions of low incentive will demonstrate greater attitude change in the direction of the position taken in the speech than persons administering positive feedback under conditions of high incentive;_conversely, (2) Persons administering negative feedback to a speaker under conditions of low incentive will demonstrate greater attitude change counter to the position taken in the speech than persons administering negative feedback under conditions of high incentive. In order to test this hypothesis, multiple comparison Eftests were performed on all possible two treatment combi- nations of the experimental groups. The results of these comparisons indicated that Hypothesis 6 is not supported; no significant differences existed among the conditions (Table 30). Discussion In contrast to Experiment I, Experiment II provides only a modicum of support for one of Bem's major theoretical premises, that on occasion an individual infers his own 87 av av av av av av av co.an n.aua.oaa av av av av av av o>.au +.aua.oua av av av av av no.0 a.aum.0sa av av av av mm.a +.aum.o-a av av av oo.a I.ana.o|m av av mm.o +.ana.0n: av mm.o: n.alm.01: ma.a +.auz.o-m vo.a| oh.a| mm.o mo.a oo.a mm.o mm.oa ma.a memo: u.aua.0ua +.aua.0ua n.aum.0ua +.aum.o:a u.aua.oum +.aua.0um u.anm.0um +.anm.o-m cOauaocoo .mmsoum amucmEauwmxm mo mmuoom mmcmno mpsuauum Game No accuaummaoo ammulm .om mam