A STUDY OF PRIOR COMM1TMENT, DISSONANCE, AND ATTITUDE CHANGE THESIS FOR THE DEGREE OF PH. D. MECHIGAN STATE UNWERSlTY RECHARD DEAN CARTER 1971 LIBRARY fly I Michigan Stan: IIII III III III II IIII IIII IIII II IIIIII IIII III III III IIII L University This is to certify that the thesis entitled A STUDY OF PRIOR COMMITMENT, DISSONANCE, AND ATTITUDE CHANGE presented by Richard Dean Carter has been accepted towards fulfillment of the requirements for Education Ph.D. degree in Date M191]..— 0-7639 Ln. A—GI-vw" ABSTRACT A STUDY OF PRIOR CWI'IMENT, DISSONANCE, AND ATTITUDE CHANGE BY Richard Dean Carter The Purpose of the Study The purpose of this investigation was to determine if a change of attitude could be induced and explained by dissonance in subjects 'with high prior commitment about an issue. The primary questions con- sidered in the study were: 1. Can a change of attitude be induced and explained by dissonance in subjects who have a high degree of prior commitment? When engaging in an act which is discrepant with one's private beliefs, does a high commitment situation (proposed-televised-publicly identified use of statements) or a low conmiUnent situation (proposed-anonymous-non video recorded use of statements) result in the subject's experiencing a greater change of attitude? Does more or less verbal justification for subjects engaging in a discrepant behavior result in greater attitude changes? Richard Dean Carter Procedures Followed in the Study A review of related literature was made which led to the con- clusion that dissonance theory and the theory related to the social judgment-involvement approach to attitude change would be of particu- lar importance in interpreting the results of this study. The literature revealed that these two major theories could lead to contrasting expectations concerning the results of the study. The subjects were 145 students at Wisconsin State university- Superior. The university, just prior to the beginning of the 1969-70 school year, had to implement a dramatic increase in non-resident tuition and fees. Although evidence led to the conclusion that most students had strong feelings in Opposition to the tuition and fee increase, non-resident and resident students were respectively desig- nated as having either high.prior commitment or low prior commitment about the issue. All the subjects, having been designated as having either high prior conmitment or low prior conmitment, were assigned either to one of the experimental groups which performed a belief discrepant behavior or to a control group where no belief discrepant behavior was performed. The experimental and control group subjects were placed in a high commitment situation (proposed-televised-publicly identified use of the statements) or in a low commitment situation (proposed-anonymous-non video recorded use of the statements). The subjects in the experimental groups were given one or two levels of verbal justification for par- ticipating in the belief discrepant behavior. SUbjects in the control situations were given one level of justification. .A semantic differential was used as the pre-test and post-test measure. Both a Richard Dean Carter multivariate analysis of covariance and a univariate analysis were used in analyzing the data. The basic design of the study was a 2 x 3 x 2 factorial design. Findings and Conclusions of the Study Significance tests were made on: 1. low prior commitment v§_high prior commitment 2. low justification v§_control and high justification v§_ control 3. low commitment v§_high commitment 4. prior commitment x_justification interaction 5. prior commitment x_commitment interaction 6. justification x_commitment interaction 7. prior commitment ijustification x_commitment interaction The multivariate analysis of covariance did not yield any statistically significant results. The univariate analysis produced some statistically significant results, but not in any consistent manner. On the basis of the statistical evidence accumulated, none of the three null hypotheses related to the investigation could be rejected. The failure of the data to support the first hypothesis of this investigation, which was related to the ability of dissonance to induce and explain a change of attitude in subjects with a high degree of prior commitment, in large confirms the findings of social judgment-involvement approach researchers. Social judgment-involvement theorists predict that with great discrepancy between one's own position and that advocated in a contrary communication or role, little or no change of attitude will occur. Hewever, a Speculative explanation Richard Dean Carter for the failure of the data to warrant rejection of the first null hypothesis can be found in dissonance theory. As dissonance can be reduced by means other than a change of attitude, it can be suggested that an alternate form of dissonance reduction was chosen by the subjects. A review of some of the experimental subjects' Speech outlines indicated that some subjects did reduce dissonance by slanting their outlines in such a manner as to reduce their dissonant situation. Since no main effects or clear interaction effects related to either commitment or justification were found, the second and third null hypotheses of this investigation were not rejected. No generalizations related to the hypotheses appear valid from this investigation. Although the data does support the position taken by social judgment-involvement theorists, alternative dissonance explanations are plausible and further research is suggested. Implications for Further Research On the basis of the limitations and conclusions of this study, research in the following specific areas is warranted: l. A.further investigation, with apprOpriate controls for alternate modes of dissonance reduction, to determine if dissonance can induce and can explain attitude change in subjects with what has been labeled as a high degree of prior commitment in this study. 2. An investigation to determine the nature of verbal justification as perceived by the experimental subject. 3. .A thorough study of the impact of role playing on producing long term.changes of attitude. A STUDY OF PRIOR COB/MITMENT, DISSONANCE, AND ATTITUDE CHANGE By Richard Dean Carter A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Administration and Higher Education 1971 ACKNOWLEDGEMENTS This study is the result of splendid cooperation and interest on the part of many persons who gave so liberally of their time and energy. Special acknowledgement is due the members of my advisory committee, Dr. Van C. Johnson, Chairman, Dr. Robert Craig, Dr. William Ewens, and Dr. Richard Featherstone, whose patience and interested guidance have assisted greatly in the completion of my degree. Appreciation is given to the staff and students at Wisconsin State University-Superior, who unselfishly donated their time to this study. Acknowledgement is given to Mrs. Rosemary Cronk, Dr. Rhea Das, and Mrs. Linda Ilminen, whose respective assistance in editing, interpreting of data, and typing of this thesis were immeasurable. Very special appreciation is expressed to my wife, Mary Ann, and to our children, Mary Margaret, Teresa, Stewart, and Mark, whose sacrifices and patient understanding allowed for the completion of this degree. A note of thanks is due my parents, Basil and Melba, whose faith in me has always been a source of inspiration. ii TABLE OF CONTENTS LIST OF TABLES ......................... CHAPTER I. INTRODUCTION ...................... Statement of the Problem .............. Need for the Study . ................ Theory and Definition of Basic Terms ........ Organization of the Dissertation .......... II. REVIEW OF RELATED LITERATURE. . ............ Review of Representative Literature Related to the Theory of Cognitive Dissonance ........... Review of Representative Literature Related to the Social Judgment-Involvement Approach ........ Contrasts in Dissonance Research and Research Related to the Social Judgment-Involvement Approach. An Overview of Literature Pertinent to this Study. . III. PLAN OF PROCEDURE . . . . . .............. Description of the Population ............ The Experimental Design and.Methodology ....... Description of the Measuring Instrument ....... Hypotheses Used in this Investigation ........ Statistical Techniques Used for the Analysis of the Data .......... . . . . . ......... Summary. . . .................... IV. ANALYSIS OF DATA. . . . . . .............. ApprOpriativeness of the Instrument ......... Tests of Significance ................ The Effects of Commitment ............ The Effects of Justification ........... Interaction Effects ............... Discussion ..................... .A Summary ...................... iii Page v CD-hNN to 13 17 25 28 28 31 35 41 42 44 46 47 SO 53 S4 54 S7 60 CHAPTER Page V. SUlIMARY AND CONCLUSIONS ................ 63 Procedures of the Study ............... 64 Summary of Findings ................. 65 Conclusions ..................... 6S Implications for Further Research .......... 67 LIST OF REFERENCES ....................... 68 APPENDIX A The Semantic Differential Employed in this Investiga- tion . . . ..................... 72 APPENDIX B Directions Given to the Experimental and Control Groups ....................... 80 .APPENDIX C A.Description of the Fortran IV Program Developed by Jeremy Finn and Medified by David wright ...... 89 iv TABLE 10. 11. 12. 13. LIST OF TABLES The Population Used in the Investigation .......... Within Cells Correlation Listed by Corresponding Semantic Differential Scales . . . . ................ Standardized Regression Coefficients-~Independent by Dependent Variables . . . . . . . . . .......... Matrix of Correlations after Adjusting for the Effects of the Co—Variables with Related Variance and Standing Deviation . . . . . . . ...... . ........... Regression Analysis with the Six Pre-Tests as Covariates. Canonical Correlations and the Percent of Variation in DependentVariables............... ..... Cell Means for Both Post-Test and Pre-Test Results ..... Summary Table of F Ratios--Listed by Experimental Condition for Miltivariate Tests of Equality of Mean Vectors . . . . . ...... . ........ . ..... Summary Table of Analysis of Low Prior Commitment vs High PriorCommitment. . . . . . . ............... Summary Table of Analysis of Low Commitment v_s_ High Commitment . . ...... . ................ Summary Table of Analysis of Low Justification vs Control and High Justification v_s_ Control . ..... . ...... Summary Table of Analysis of Prior Commitment x Justifica- tion Interaction. . . . . . . . ........ . . . . . . Summary Table of Analysis of Prior Commitment x Commitment Interaction . . ......... . . . . . ...... . . Page 31 39 40 47 48 49 51 52 $3 54 SS 55 56 TABLE Page 14. Summary Table of Analysis of Justification x_Commitment Interaction ....................... 56 15. Summary Table of Analysis of Prior Commitment x Justification x_Commitment Interaction. . . . 7' ..... 57 vi CHAPTER I INTRODUCTION It has been demonstrated that private Opinion may be changed when an individual finds it necessary to express a public position contrary to his private attitudes. The reasons for involvement in situations of such discrepant nature vary widely. The effect such situations have on one's commitment to behavioral actions is important to this investigation. There are several theoretical positions which attempt to explain the degree of attitude change that can reasonably be expected when an individual finds himself in a position of expressing public opinions contrary to his private attitudes. Sherif's social judgment-involvement approach and Festinger's theory of cognitive dissonance are two of the major theories related to this problem. These two theories lead to different predictions concerning the potential attitude change. As McGuire indicated, a degree of theoretical controversy centers around the discrepancy-involvement question.1 This investigation will attempt, in part, to extend the research related to the discrepancy-involvement question in the study of the change of attitudes. 1William J. McGuire, "The Current Status of Cognitive Consistency Theories," in Shel Feldman (ed.) ngpitive Consistency: Motivational Antecedents and Behavioral Consequents (New York: Academic Press, 1966), p. 19. I. Statement of the Problem The purpose of this investigation was to study if a change of attitude could be induced and explained by dissonance in subjects with high prior commitment about an issue. The primary questions considered in the study were: 1. Can a change of attitude be induced and explained by dissonance in subjects who have a high degree of prior commitment? When engaging in an act which is discrepant with one's private beliefs, does a high commitment situation (proposed-televised-publicly identified use of statements) or a low commitment situation (proposed-anonymous-non video recorded use of statements) result in the subjects having a greater change of attitude? Does more or less verbal justification for subjects engaging in a discrepant behavior result in greater attitude change? II. Need for the Study A number of recent studies differentially lend support to two of the theories related to the study of attitude change. Both approaches, most commonly associated with the theories of Festinger and Sherif, evaluated.what takes place when the individual confronts or creates a communication discrepant with his own attitudes. Festinger's theory of dissonance predicted that the larger the discrepancy between a person'a private position and a discrepant communication or action, the greater will be the resulting attitude change. To the contrary, Sherif's social judgment-involvement approach predicted that when small discrepancies exist a change of attitude may occur, but large discrep- ancies with one's private position will result in little or no attitude change.1 Involvement of subjects with an issue is viewed differently in the two theories. Sherif indicated that apparent contradictions between the social judgment-involvement approach and dissonance theory are in.part explained by the failure of dissonance theorists to con- sider the involvement of subjects in the issue. .As Rand indicated, the concept most related to involvement in dissonance theory is how important 2 The social judgment-involvement the issue is to the individual. approach leads the investigator to expect less attitude change with the greater involvement of the subjects. On the other hand, dissonance theory leads the investigator to conclude greater change is more prdbable as the importance of the issue to the individual increases. These apparent contradictions indicated a need for further exploration of the effect of discrepant communication or discrepant action on the change of individual attitudes. The need for further exploration concerning individual involvement and attitude change is also evident. This investigation is, in part, an attempt to meet this need for further research. 1MelvinA. Rand, "An Empirical Comparison of Sherif's Social Judgment Approach and Festinger's Dissonance Theory at Their Points of Contrast Ego Involvement and Discrepancy of Communication," (unpublished Doctoral dissertation, University of Oklahoma, 1967), Dissertation Abstracts, 1967, 29 A, l: 1128-1129eA. 2125:.- at.- III. Theory and Definition of Basic Terms .MTNM'// The basic background of the theory of cognitive dissonance consists of the idea that the human organism.tries to establish internal harmony or consistency among his opinions, attitudes, knowledge, andvalues.1 The term cognition refers to any knowledge, / I I behavior.2 Dissonance is used to represent inconsistency between cog- I I opinion, or belief about the environment, one's self, or one's nitive elements. Two cognitive elements are dissonant if they do not agree or if, for one reason or another, they do not fit together1§,»»/< FOr example, if a person knew he could only afford.to pay $250 for a new home appliance and he had just been persuaded to sign a contract for an appliance costing $400, there would be a dissonant relation between these two cognitive elements. To the contrary, two cognitive elements are consonant if one follows from.the other. Thus, the knowl- edge that it is twenty degrees below zero is consonant with a chilled feeling when you leave the house to go outside. Consonance and dissonance may exist only between relevant-fimlv I» elements; therefore, if the elements are not related, there is no basis I fer dissonance.4 The degree of dissonance is determined by the propor- I V" tion of all relevant cognitive elements that are dissonant. Secord and I I l I Backman explained: I The magnitude of dissonance is a function of the proportion I of all relevant cognitive elements that is dissonant. These elements I 1Leon Festinger, A Theory of Cognitive Dissonance (Stanford, California: Stanford university Press, 1957), p. 260. 2 Ibid., p. 3. 31bid., p. 12. 4Ibid., pp. 11-13. \/ are generally weighted according to their importance. Thus, the magnitude of dissonance may be expressed in terms of the following ratio: e = importance X no. of dissonant elements Dissonanc l. importance X no . of consonant elements From this ratio it is clear that the more nearly equal the relative proportions of consonant and dissonant elements, the greater the dissonance is. If there are only a few dissonant elements and.many consonant elements, dissonance is relatively low. The number of dissonant elements can never exceed the number of consonant eIements, for this would lead to a change, removing the dissonance. The ratio is used to explain the concept and is not intended as a.measuring device.2 Hewever, the magnitude of dissonance created can also be viewed as the magnitude of consequent attitude change.3 According to Festinger, behavior change is an attempt to reduce dissonance. The strength of the psychological pressure to change behavior is a function of the magnitude of dissonance.4 There are some interesting implications of dissonance theory that are of special importance to this investigation. First, the:more important the belief is to the individual, the greater will be the magnitude of dissonance if the individual is led to engage in a behavior discrepant with that belief. Secondly, whenever a person's cognitions about external forces lead to a public stand different from.the privately held attitude, dissonance will result.5 Following the logic 1Paul F. Secord and Carl W. Backman, Social Psychologx (New York: McGraw-Hill Book Company, 1964), p. 116. 21bid., p. 116. 3Arthur R. Cohen, HAttitudinal Consequences of Induced Discrepancies Between Cognitions and Behavior," Public Opinion Quarterly 1960, 4' 297. 4Festinger, 92, 913,, p. 18. SCohen, pp. 513., pp. 297-298. of the theory, increasing the justification for taking a public position contrary to one's private attitude will reduce the magnitude of dissonance and corresponding attitude change. That is to say, the minimal justification that will elicit the dissonant behavior will result in the largest magnitude of dissonance and the greatest change. The logic of the theory further indicates that in forced compliance situations, such as this investigation, the least justification that 'will elicit the behavior contrary to one's attitude will result in the greatest change of attitude.1 There are some theoretical differences between the theory of cognitive dissonance and the social judgment-involvement approach to attitude and attitude change which relate to this investigation. The theory of the social judgment-involvement approach will be reviewed primarily at the points of contrast with dissonance theory. It can be postulated from.the social judgment-involvement approach that if a subject is highly involved with his position, a large discrepancy between his personal stand and a communication or public action will rarely result in a change of attitude. Sherif and HOvland stated: With small discrepancies between the position of connnmication and that of the subject, changes of Opinion in the direction advocated by communication will ocOur. 'With large distances between the stands taken in communication and by the subject, Opinion changes will be infrequent. Changes in the direction opposite to that advocated by commmication will be more frequent when the discrepancy between the stands taken in c 'cation and by the subject is large.2 1Festinger, pp, git,, pp. 260-263. ZMizafer Sherif and Carl 1. Hovland, Social Jud nt: Assimilation and Contrast Effects in Communication.and* ttitude Change (New'Haven: YaleUhiversityPrjfs, 1961), p. 157} \/ An individual's stand is conceived as an internal anchor in the social judgment-involvement approach. Any discrepancy between the individual's stand and a particular communication is considered a relative difference to be defined by an internal reference scale for rejection, acceptance, and noncommitment. Ego involvement is con- sidered to be of prime importance in understanding the reference scales of the individual. An assimilation range is also a part of the social judgment-involvement approach to attitude change. This range refers to the individual's latitude of acceptance of positions which are not highly objectional to himt This is to say that the less the individual is involved in his stand, the greater will be his assimilation range. The importance of the assimilation range and personal involvement in an issue to understanding the implications of the theory are illustrat the following statement of Sherif, Sherif, and Nebergall: To sum up so far, there is now evidence that increased discrepancy between.initia1 position and the communication does increase the likelihood of change up to a certain point and that this point is affected by the degree of involvement in own stand. When involvement is high, increasing discrepancies lead to unfavor- able evaluation of the communication, contrasted placement of its position, and decreased tendency to change toward the communication, with boomerangs away from the position advocated when.both discrepancy and eg9_involvement are very great.I e social judgment-involvement approach and the theory of cognitive dissonance lead to different predictions related to the hypotheses of this investigation. The synopses of the two theories were presented from the standpoint of their potential to assist in the interpretation of the results of this investigation and should not be considered as a comprehensive review of either the social judgment- involvement approach or the theory of cognitive dissonance. 1Carolyn'W. Sherif, .Muzafer Sherif, and Roger B. Nebergall, .Attitude and.Attitude BChfififi—i The Social Judgment- -Involvement Approach IPhiIadelphia: ecf'Company, 1965), pp. 196- 197. ' IV. Organization of the Dissertation The remainder of this dissertation has been divided into four chapters. Chapter II is concerned with a review of the literature pertinent to this investigation and is divided into four main areas: 1. A.review of representative literature related to the theory of cognitive dissonance. 2. .A review of representative literature related to the social judgment-involvement approach. 3. Contrasts in dissonance research and research related to the social judgment-involvement approach. 4. An overview of literature pertinent to this study. Chapter III is devoted to the plan of procedure followed in this study. In the major sections of this chapter, consideration.will be given to the description of the population, the experimental design, description of the measuring instrument, the hypothesis, statistical techniques for analysis of the data, and a chapter review. In Chapter IV an analysis of the data is presented. Chapter V is devoted to the summary and conclusions of the dissertation. CHAPTER II REVIEW OF RELATED LITERATURE Much of the literature related to attitude change has emerged in the last two decades. Two of the major theories articulated in the past twenty years were Festinger's theory of cognitive dissonance and Sherif's social judgment-involvement approach (assimilation contrast theory), both of which are discussed in this chapter. In addition, literature related to the points of difference between these two theories will be reviewed. I. Review of Representative Literature Related to the Theory of Cognitive Dissonance In self persuasion experiments a common occurrence has been the change of attitudes by individuals. The theory of cognitive dissonance has frequently been used to explain this phenomenon. For example, it can be indicated that whenever an individual behaves in a manner discrepant with his private attitude, dissonance may result; one way to reduce the resultant dissonance is to change the attitude. Gross predictions about the degree of attitude change can.be made on the basis of dissonance theory if consideration is given to the importance of the issue to the individual (prior commitment), commit- ment, and justification for engaging in the discrepant behavior. The magnitude of dissonance increases as the importance of the Opinion to the person increases.1 Research by Cohen supported the 1Festinger, op, 913,, p. 263. \// 9 \I concept that prior commitment (importance of the opinion) will in part control the magnitude of dissonance: The amount of dissonance will increase as the general importance of the guiding cognitions increases , and as the balance between the Opposing cognitions increasegj When a ~p‘eI'son'Te‘x'pOffer—1E”dissonance, he will try to reduce or eliminate it in prOportion to its strength. In general, he may do this by making his cognitions more consistent with the expression already made. Thus, one major consequence will be a change in one's privateopinion to make it more consistent with the position expressed. The magnitude of (11.552119999115. effectedflbwdegree o M..- commitment-~pub1ic versus private commitment or identified versus _fi fl.“ MM anonymous statements. Normally we do not know if a commitment has .o-v' been made unless it is public; however, it can be assumed that public commitment is usually more resistant to change than private commitment. A further assumption may be made: resistance to change will increase as the strength of expectation the individual will hold to the commitment increases . 2 A 1968 study by Helmreich and Collins investigated implications of dissonance theory under three levels of commitment. These three levels of commitment were identified as anonymous audio recording, iden- tified video recording with subsequent explanation, or identified video recording with no opportunity to recent. In the Helmreich-Collins study subjects were enticed to record a strongly counter attitudinal statement about an important issue. Subjects were paid either $.50 or $2.50 for the counter attitudinal task. In the two higher commitment conditions subjects given low financial inducement showed significantly more 1Cohen, g1. 21in p. 293. 2Jack W. Brehm and Arthur R. Cohen, Explorations in Cognitive Dissonance (New York: John Wiley and Sons, Inc., 1962), p. 8. / 11 attitude change than those given the larger sum. In the low- commitment condition the larger amount of money produced more attitude I change. LIKE/commitment effect was significant at the .05 level and according to the authOrs, in large, the study found dissonance effects s.._,.,'——-~ on the commitment issue.1 The amount of justification like the degree of commitment plays an important part in predicting attitude change according to dissonance theorists. The effect of justification is illustrated in the Secord and Backman summary of the often-quoted 1959 experiment of Festinger and Carlsmith. In the Festinger and Carlsmith experiment subjects were told after the completion of a dull, boring experimental task that a student helper usually brought in the next eXperimental subject [<3 and told him how enjoyable the experiment was. It was implied the III person who normally completed this function had not shown up. The KS subject who had just completed the dull and boring task was asked to A I? fill in for the missing helper. Two degrees of reward (justification) “(J for serving in this role were given. Some subjects received $1 for their efforts, while other subjects received $20.2 After the subjects had served as helpers, they were interviewed and asked to rate their opinions concerning the experiment on an eleven- point scale. The scale ranged from maximum negative opinion to lRobert Helmreich and Barry E. Collins, "Studies in Forced Compliance: Commitment andeagnitude of Inducement to Comply as Determinants of Opinion Change," Journal of Personality and Social Psychology, November, 1968, 10: 75—81. 2Secord and Backman, op, git,, p. 118. \/ 12 maximum positive opinion.1 .As would be predicted from dissonance theory, as the amount of justification (dollars) increased, the weaker was the tendency to change private opinion to correspond with.what the subject had said in the experiment.2'/As justification increases IfOTflEthgihg‘ih—EEE—diggrgpant behaviOr: the magnitude of dissonance decreases; consequently, the degree of expected attitude change decreases. The results of this experiment of Festinger and Carlsmith has been a point of controversy between incentive and dissonance theorists. These differing interpretations will be discussed in a later subsection‘of_this'dissertatign;/ Further evidence in support of the dissonance theory interpretation comes from a subsequent SthYQPXEQQEEE;) In the Cohen experiment college students were paid either $.50, $1, $5, or $10 to ///;77 write a counter-attitudinal essay. The resulting data showed that the less the justificationflLdollars) for engaging in the discrepant behaviOr, the greater the attitude change.3 BTW” Aronson found dissonance results in an experiment with toy preference. In individual sessions, he asked each child to rank the attractiveness of several toys; then he indicated to the child that he could not play with the toy he had considered very attractive. Degrees of punishment were threatened for transgression. .Aronson found that lIbid. , p. 118. 2Cohen, 92, 913;, p. 299. $Brehm.and Cohen, Op. cit}, pp. 73-78. J 13 children devalued the forbidden toy only if forbidden to do so under mild punishment. The threat of severe punishment did not yield similar results.1 I Shmilar dissonance results have been foundIby several researchers. However, incentive theorists and assimilation contrast theorists question the results of some of these experiments and have conducted experiments which support their theoretical position. K II. Review of Representative Literature Related to the Social Judgment-Involvement Approach Assimilation contrast theorists, such as Sherif, contend that the greater the dedication of an individual to his stand on an issue (prior commitment), the less likely he is to change that stand as a result of a discrepant communication or action. In,a communications study related to the 1960 presidential election, Sherif and his fellow researchers found evidence that highly involved subjects changed their attitudes less frequently than those with less personal involvementt? I (N... M. Also, there was evidence of the fact that high ego-involvement -' ~ ”‘1” «". ‘ restricts the individual's asshmilation range ' Thus, the mgge highly involved thekipdividual_iswith the issue the less the number of MN ‘ ~ ’DRN/«M counter-attitudinal arguments will be acceptable to him. 1Elliot Aronson, "The Psychology of Insufficient Justification: An Analysis of Some Conflicting Data", in Shel Feldman (ed.) Co itive Consistency: IMotivational Antecedents and BehaviOral COnsquefits, 92. gig, p. 1I2. \IW/ Z‘Shefif’ Sherif’ and‘Nebergall’ _2- 2.1.15. p. 226. J ‘3 "We . . . (ggarolyp W} Shegif andIMuzpfer Sherif "The Social Judgment- Involvement Approach to Attitude," in Carolyn‘W. Sherif and.Muzafer Sherif (eds.) Attitude, Ego-Involvement,and Changg (New York: JOhn Wiley and Sons, Inc., 1967), p. 152. PLEASE NOTE: Some Pages have indistinct print. Filmed as received. UNIVERSITY MICROFILMS 14 The logic of common sense would lead to a conclusion that a communication advocating a stand substantially different from one's own personal beliefs would lead to a rejection of that communication. Whittaker has reviewed such results based on WOrld war II propaganda studies. In one particular instance propaganda leaflets containing . photographs of conditions in a Prisoner of war camp in Canada were to be drOpped behind enemy lines. Before dropping the leaflets, Army personnel decided to pretest the study with German prisoners in POW'camps in Italy. The stories and photographs in the leaflet described con- ditions which actually existed in the Canadian.camp. One of the descriptions in the leaflet described a breakfast of eggs, toast, and coffee. The prisoners on.whom the leaflets were pretested had been led to believe by their German superiors that conditions in all Allied prison camps were deplorable. Although they had discovered the relatively good conditions in.their own camp, they could not accept the leaflet as truthful because it was too divergent from their prior committed position.1 As Whittaker explains, most psychological warfare activities do not qualify as scientific experiments. Yet, he has accumulated evidence in scientific studies which shows that as the person's own position becomes more extreme, the less tolerant he is of the opposing Opinions of others. In one experiment farmers were asked to choose from among nine statements concerning government controls in agriculture-- 1James O. Whittaker, "The Resolution of the Communication Discrepancy Issue," in Carolyn W. Sherif and.Muzafer Sherif (eds.) Attitude, Ego-Involvement, and Change, Ibid., pp. 159-160. §;T\ I \qu 15 . those they found Objectionable, those they found unobjectionable, and also, the one statement most representative of their own.position. The results of the experiment led to the conclusion that a small W discrepancy between the communication.and the individual's own stand K W yields small Opinion change. Moderate discrepancies yield either egative change from.the communication, assuming 1 negligible change 0 the i -' 1dual had a high degree of prior commitment. Experiments have shown that enacting a role which supports a (\O contrary position to one's own is more effective in changing a person's opinion than just listening to the information passively or a; __EEEEEIE§_EEEE£§_Pl§X.359—Lgl§: From the standpoint of dissonance QCILf] theory, it is possible to explain.the purported change in the individual's position through the creation of psychological tension (dissonance). ! The dissonahce can be reduced by Changinggtgwagdgthewposition advocated in the role; .However, assimilation-contrast theorists frequently U explain the attitude change from the standpoint of self-persuasion QII aspects of role playing. As Sherif, Sherif and Nebergall pointed out, role playing is usually referred to as commitment in self-persuasion experiments. The attitude change is explained by the individual becoming situationally ego-involved in doing his task well, although he may not agree with what he is saying.2 I The importance of ego-involvement in understanding the potential attitude change as a result of role playing is also explained by Sherif, Sherif, and Nebergall: When agreeing to perform a task, even a mildly unpleasant one, the individual is situationally involved in perfOrmung well, 11bid., pp. 163-176. 2Sherif, Sherif, and NebergalIC/pp, SEE): p. 197. 16 hence his differential attention to arguments that will help him. I do so. On the other hand, to the extent that he is ego-involved in his stand on the issue, the tendency to change toward the role should be less. Conversely, if he plays a role that suppgrts his own stand, ego involvement in the task should reinforce 15 own stand in prOportion to his involvement in it.1 ”Mr' .As the prOponents of the social judgment-involvement approach expect the context of attitude change to be different from that implied by dissonance theory in relation to prior commitment and commitment, the role of justification in attitude change is also viewed differently by the proponents of the two theories. From the stand- point of the assimilation—contraSt theory, attitude change by an individual is more probable as the justification for making that change is increased. For example, the assimilation-contrast theorists would expect generally that to the extent the source and the communicator stand high in the individual's reference groups, the greater the possibility the individual might change his attitude.2 .Assuming a person expresses an opinion.discrepant from.his. personal beliefs, the larger the incentive (justification) for taking that position, the greater the likelihood he will change his attitude in the direction of the discrepant position. The more the justifica- I tion, the more the eXpected change unless the justification involves I suspicion or other negative incentives. Hunt reviewed two studies by Scott related to these premises. In one of Scott's studies a prize of I~ $100 was offered to the winners of an elimination debate. One-third of the subjects defended a neutral position, one-third of the subjects defended their own positions, and one-third of the subjects defended a . 1/ Ibid., pp. 197-198. I 21bid., p. 16. \p 17 position opposite their own. Scott discovered that winners, regard- less of their own attitudes, moved toward the position they defended in the debate significantly more than did the losers. .Although this study does not meet the assimilation-contrast theory requirement of knowing C1" the magnitude of justification (incentive) beforehand, the theorists would postulate that the better the individual's performance, the more likely will be the chance that the individual will convince him- . .. “Willifl self of that position.1 “~*““ Neither dissonance theorists nor assimilation-contrast theorists would stress that their contrasting expectations concerning an individual's change of attitude would always differ as a result of isolating the effect of a single variable such as prior commitment, commitment, or justification. HOwever, under many circumstances any of these variables or interaction of these variables could lead dissonance theorists or assimdlation-contrast theorists to expect contradictory results in attitude change experiments. The theoretical points of contrast and contradictory findings are discussed in the following section of this work. III. Contrasts in Dissonance Research and Research Related to the Social Judgment-Involvement Approach afgvgp Brehm and Cohen mentioned the effect of the importance of the dissonant cognitive element (prior commitment) to the individual in their book, Exploration in CognitiVe Dissonance. They used as an lMartin Floyd Hunt, Jr. , "Open- And Closed-Mindedness And Self-Persuasion: Incentive vs. Dissonance Theory," (unpublished Doctoral dissertation, Michigan State university, East Lansing, Michigan, 1969), pp. 9-10. 18 example a child who is induced to eat a disliked vegetable. The greater the child's initial dislike for the vegetable, the larger will be the magnitude of dissonance. The magnitude of dissonance will also be influenced by the importance of the reward the child receives for eating the disliked vegetable.1 ’rV —-—.1____________ . . Brehm.and Cohen explain that the Child's dissonance could be reduced in numerous ways. Dissonance might be reduced by increasing the liking for the vegetable, which would increase consonance and thus reduce dissonance. He might devalue the importance of his dislike fer the vegetable, or he might reduce dissonance by magnifying the value of the reward he receives fOr engaging in the dissonant behavior.2 The expectation that the greater the degree of prior commitment, the greater the prObability of a change of attitude is in opposition to the stand taken by those using the logic of the social judgment-involvement approach. The importance or relevance of a tOpic to an individual ego- involvement is perceived by the Sherifs as an internal anchor fer the individual. The more involved the individual is in his stand, the more difficult it is to persuade him to change his position.3 Sereno V/ conducted a communications study to investigate if subjects highly involved in their own position would change their attitudes on a topic less than subjects with lower involvement in the tOpic. He also attempted to determine if highly involved subjects lowered their L/ 1Brehmand Cohen, 9p, git}, pp. 6-7. 21bid., p. 7. 3Sherif and HOvland, pp, 213,, pp. 174-175. 19 evaluation of the source more than the less involved subjects. He I concluded that a highly credible source did change attitudes of lowl£I involved subjects significantly more than those of highly involved ‘\I/// subjects; too, highly involved subjects tended to lower their evalua- I tion of the source more than lowly involved subjects. In brief, his findings SUpport the theoretical position of the Sherifs on the subjec of ego-involvement.1 . // In this investigation contrasting r 5 ts could be expected in view of the literature regarding the theorie of cognitive dissonance and the social judgment-involvement .It is possible to postulate from dissonance theory that subjects with high prior commitment would be susceptible to a larger magnitude of dissonance than subjects with a lower commitment on an issue. The larger magnitude of dissonance resulting from higher prior commitment would increase the likelihood of a subject's change of attitude if that fOrm of dissonance reduction was chosen by the individual. In contrast, on the basis of the logic used and the empirical evidence associated with the social judgment-involvement approach, it could be postulated that the more involved the subject is with his stand, the less likely he is to have a significant change of attitude. .——__.—/ Festinger, in discussing the magnitude of dissonance,j$§$tioned that as the dissonance introduced b statedldi§§8£EETEEE_‘h‘others decreases, the magnitude of dissonance decreases. In other words, con- sonance can be increased with the knowledge that other people hold the 1Kenneth K. Sereno, "Ego-Involvement, High Source Credibility and Response to a Belief-Discrepant Communication," Speech Monographs, November, 1968, 35: 476-481. 20 same Opinion.1 In this investigation higher commitment connotes increased public exposure of the position taken by thefisubjeetL/ If t e su ject is taking a position contrary to his private belief, it could be expected that the magnitude of dissonance may increase as the degree of public exposure increases. On the basis of literature related to dissonance theory, it would be expected that more attitude change would occur when subjects with the most extreme position play a role Opposite their * stand. This occurrence would be explained by the fact that this situation.would create the greatest discrepancy for the individual and thus the greatest dissonance)\ The social judgment-involvement approach, on the other hand, would lead to the conclusion that the more the subject is ego-involved in his stand, the less the tendency to change his attitude toward the role he is playingI The social judgment-involvement prediction was a§§§§§§d in a tudy beEIbihg\ and_EepoEted_bygShegifi_§he:if,_apdgyebepgaII. .Althoughqgflang found signs of a dissonance\effect, heJconcludedgthat,\gontrary:togdissonance theory, there was no tendency for subjects with an extreme stand in a role playing situation to change their attitudes more than subjects with.moderate stands. Elbing found a trend in the Opposite direction, that is, moderates had more attitude change. His findings support the position advocated in the social judgment-involvement approach.2 In a study of the interaction of commitment and dissonance, Kiesler, Pallak, and Kanouse found some typical dissonance effects, ‘ 1Festinger, pp, 913,, p. 263. 2Sherif, Sherif, and Neberga 1, pp, 313,, p. 198-199. 21 as well as compartmentalization of issues. In their study commitment to consonant behavior was varied orthogonally to the degree of I. dissonance produced by a logically related later behavior High commitment was identified as public use of a subject's speech; low commitment was private use of part of the subject's speech with con- ditions allowing for complete anonymity. Dissonance was manipulated X by the degree of choice in writing a counter-attitudinal essay. The .1-._..._,_ m-Wk M results of the experiment showed that high-dissonance subjects changed ,___+M more than low-dissonance subjects on the dissonance topic. On the related commitment tOpic, high dissonance subjects changed more than low only under low or no prior commitment. The experimenters felt that I ' 'tment subjects compartmentalized the two issues.1 The relationship of justification and dissonance. is widely discussed in - literature. A representative 5 y of this tOpic is the Festinger and Mrlsmith study, which are. ed a great deal of interest and response, both pro : d con, in the . terature. After completing a dull and boring experimenta task, i e subject was told that a student assistant usually brought in ' .~ next subject and told him how much fun the experiment was. I was in: 'cated that the assistant had not arrived. The student . O had just c- n eted the dull and boring task was asked to fill ' for the student assis -. t. T‘wo magnitudes of justification, a reward of $1 or $20, were offe d to the subject for 1Charles A. Kiesler, Michael S. Pallak, and David E. Kanouse, "Interactive Effects of Commitment and Dissonance," Journal of Personality and Social Psycholggy, vol. 8, No. 4 (Marc-:5, 19685 pp. 331-338. 22 serving in this role. It was predicted that performance of such a belief-discrepant behavior would produce dissonance. The subjects who were given relatively low justification would experience more dissonance than subjects given the high justification. The dissonance theory position was supported as predicted; those subjects who received only $1 rated the experiment significantly higher on the post-test ratings than did the subjects receiving $20.1 Festinger theorized that the larger the pressure used to elicit the overt behavior, the weaker would be the tendency for the subject to change his private opinion to correspond with what he had done or said. This assumed the pressure was beyond the minimum to elicit the behavior. The results of this study were in line with Festinger and Carlsmith's prediction, but an alternate interpretation of the data was feasible on the basis that subjects paid the $1 condition worked harder at telling the interviewer that the tasks were fun and enjoyable. To determine the plausibility of such an alternate explanation, the experimenters taped each subject in an interview situation. The interviews were than transcribed and rated by two independent raters on five dimensions. The correlations between the two independent raters ranged from .61 to .88. The difference between the mean ratings for $1 and $20 subjects was not significant. The experimenters concluded that subjects paid $1 did not improvise more convincingly than those paid $20. An alternative to the dissonance explanation of the attitude change reported in subjects paid $1 did not seem reasonable, and Festinger and Carlsmith's predictions were supported.2 1LeonFestinger and J..M; Carlsmith, "Cognitive Consequences of Forced Compliance," Journal of Abnormal and Social Psychology, 1959, 58: 203-210. ,» 2Ibid., pp. 208-210. V// 23 The Festinger and Carlsmith study has been reviewed and critiqued by several peOple. Rosenberg, Chapanis and Chapanis, and Janis and Gilmore--all indicated concern that the large monetary incentive given for the counter-attitudinal performance may create suspicion on the part of the subject. Speaking of the dissonance studies, Rosenberg noted: It is only because that research has utilized strong manipulations, which many subjects may find Odd or extremely surprising, that some of the critical response to these studies has focused upon the possibility that the main effects that they disclose could be artifactual in nature, or could be due to certain special and unsuspected dynamics whose clarifici- tion might limit the proper sc0pe of dissonance theory itself. Janis and Gilmore, using a 2 x 2 x 2 factorial design, studied the independent variables of unfavorable versus favorable sponsorship, small versus large monetary reward, and overt role playing versus passive eXposure. They ascertained the main effects and inter— action of the three independent variables and reported the following: . Analysis of variance of attitude scores showed that there were no significant main effects but, as predicted by incentive theory, there was a significant interaction such that more attitude change occurred when overt role playing was carried out under favorable sponsorship than under unfavorable g sponsorship conditions. In general, the results fail to verify predictions from dissonance theory apd tend to support an incentive theory of attitude change. Aronson disputed the conclusions reached by Janis and Gilmore, as well as the conclusions reached by Rosenberg regarding the EMilton J. Rosenberg, "Some Limits of Dissonance. Toward a ' -‘ iated View of CountersAttitudinal Performance," in Shel Feldman (ed. ) Qggnitive Consistency. Mbtivational Antecedents and Behavioral Consequents, Op, Cit. p. 143. 2Irving L. Janis and J. Bernar' Gilmore, "The Influence of Incentive Conditions on the Success 0 Role Playing in.MOdifying .Attitudes," JOurnal Of Personality d Social Psychology, vol. 1, No. 1, 1965, p. 17. / 24 Festinger and Carlsmith study. Aronson pointed out that under close inspection Janis and Gilmore's conditions do not correspond with those used by Festinger and Carlsmith. In particular, Janis and Gilmore did not indicate to their subjects that they would be on call as part of their obligation for receiving the agreed-upon renumeration. Aronson felt that the Janis and Gilmore procedure would arouse more suspicion than the Festinger and Carlsmith study. In brief, Aronson agreed that role playing may have an effect beyond that predictable from dissonance theory but maintained that the Janis and Gilmore study does not refute the conclusions of the Festinger and Carlsmith study nor provide an "unambiguous" test of either dissonance or incentive theory. In a similar fashion, he challenged Rosenberg's criticisms of the Festinger and Carlsmith study. Fprther, he indicated that a study by Arthur Cohen using sums of $.50, $1, $5, and $10 met much of Rosenberg's criticism.and demonstrated the dissonance position.l xxfiesWfi inverse relationship between justification and attitude change as suggested by dissonance theory. Studies by incentive theorists and those using the logic of the social judgment-involvement approach to attitude change, such as Sherif and Sherif, Elbing, Nebergall, and others, yielded results contrasting to those suggested by dissonance theory. .Although there has been extensive analysis and interpretation \ of the data from these and other studies, there is not a large "\ (\ I 1Elliot Aronson, "The Psychology of Insufficient Justification: An.Analysis of'Some Conflicting Data," in Shel Feldman (ed.) Co 'tive Consistency: Motivational Antecedents adeBehavioral onseguents, op, cit., pp. 10:;I;1. , ,\ i \ | l 25 degree of common agreement among the various schools of thought related to the effect of justification on attitude change. \I IV. An Overview of Literature Pertinent to this Study A.number of recent studies supported two or more approaches to the study of attitude change. The approach postulated in dissonance theory and the position advocated in the social judgment—involvement approach both evaluated what takes place when the individual is presented infonmation which is discrepant from his initial position. Individuals representing these two schools of thought acknowledged that role playing may have greater influence on attitude change than merely receiving passive information. Role playing in this study is related to commitment. Also involved in this study are the variables of prior commitment and justification. Prior’eemmitment”in this study is defined a he de ree — personal involvement or importance of the issue to the individual. NM r-- In dissonance theory a greater change of attitude is predicted the more discrepant the communication or role playing situation is from the individual's private position. As reported by Rand, the social judgment-involvement approach of Sherif predicts that with great discrepancy between one's own position and that advocated in a con- trary communication or role, little or no change of attitude would occur. Small or moderate discrepancies between a person's initial position and that advocated in a discrepant communication or role 'will more likely result in a change of attitude.1 lRand, 9p. _c_i_1_:_., p. 1128-A. 26 Subjects under high commitment in this study will be under the assumption that they are to outline and later record an identified speech supporting high non-resident tuition for state universities, to be played over closed-circuit television, the local educational television station, and possibly other locations. Subjects in a low commitment situation are told that they may be asked to record their speech on the same topic in an anonymous situation oVer radio. As further assurance to anonymity, it was indicated that only certain statements from their speeches would be selected for broadcasting. From the framework of dissonance theory it would be expected that subjects with extreme positions playing a role Opposite their stand would feel a significant magnitude of dissonance. Using the framework the social judgment-involvement approach to attitude change, Sherif, Sherif, and Nebergall indicated that Elbing did not find a tendency for subjects with extreme stands to change more in'a role playing situation than subjects with.moderate stands. To the contrary, he found that moderate subjects changed more in role playing situations than extreme subjects.1 Dissonance theory framework led to the prediction of an inverse relationship between the amount of justification for engaging in a belief-discrepant behavior and the corresponding magnitude of dissonance. Festinger stated: The magnitude of the dissonance resulting from an attempt to elicit forced compliance is greatest if the promised reward or threatened punishment is either just 1Sherif, Sherif, and Nebergall, pp, SEE}, pp. 197-199. 27 If /’ / sufficient to elicit the overt behavior or is just barely not sufficient to elicit it. From the framework of the social judgment—involvement approach to attitude change, it would be expected that to the extent the greater the justification given for engaging in a belief discrepant behavior, the greater the possibility the individual would change his attitude. —. MCGuire, in discussing the discrepancy-involvement question, stated: Another persuasion area in which consistency theories have been applied with a stimulating degree of theoretical con- troversy and an attractive amount of empirical ambiguity has been that of the discrepancy-involvement question. How does the amount of attained opinion change vary as a function of the amount of change urged, and how does this relationship vary as the believer's "involvement" increases? There seem to be opposite predictions on the basis of dissonance theory (Zimbardo. . .) and of assimilation-contrast theory (M. Sherif and HOvland, . . . C. W; Sherif et. al. . . .). Zimbardo. . . has proposed 3 plausible reconciliation, but the issue is still in doubt. . . .MCGuire's statement would be applicable to this study. Neither dissonance theory nor assimilation-contrast theory can be ignored in this study, and both theories will be of use in interpreting the results of this investigation. . I lFestinger, pp. p_i_1_:_., pp. 262-263. 2William.J.McGuire, "The Current Status of Cognitive Consistency Theories," in Shel Feldman (ed.) Cognitive Consistency: thivational Antecedents and Behavioral Consequents, pp. pip. , p. 19. J CHAPTER III PLAN OF PROCEDURE The purpose of this investigation was to study if a change of attitude could be induced and explained by dissonance in subjects ‘with high prior commitment about an issue. This chapter is a description of the procedures followed in this study and includes a description of the population, the experimental design and methodology, the measuring instrument, a statement of hypotheses, a description of the statistical techniques used for analysis of the data, and a summary. I. Description of the POpulation The subjects in this investigation were all students at Wisconsin State University-Superior. This University is the smallest in the Wisconsin State University System, but it has a prOportionally large graduate school. In the 1968-69 school year approximately one- fourth of the students were non—residents. Prior to the 1969-70 school year, tuition and fees for non-resident students in the Wisconsin State University System were greatly increased. During the 1968-69 school year, non-resident tuition and fees for the full-time students at Superior were $378 for the undergraduate students and $412 for the graduate students. undergraduate non-resident tuition and fees for 1969-70 were $655, while graduate non-resident tuition and fees were $786. The guidelines for the 1969-70 fee structure 28 29 were not received by the University in time to alert most non- resident students of the fee changes prior to the students' return to campus. .At the time this experiment was conducted, students were aware of a further proposed increase in tuition and fees for the 1970-71 school year, as well as an increase in summer session fees for full-time non-resident students from a 1969 level of $202 for undergraduates and $204 fOr graduate students to a 1970 level of $330 for undergraduates and $405 for graduate students. On the basis of public announcements, informal interviews, and.newspaper editorials, it was decided that most students on the Superior campus had strong opposition to the prOportionately large increase in non-resident tuition and fees. For the purpose of this investigation, non-resident students were arbitrarily designated as having high prior commitment about this issue. Resident students were designated as having low prior commitment. It was known that most students had strong feelings about the issue, but it was reason- able to expect that non-residents were more affected.by the tuition and fee increase and should be designated as a separate group from the resident students. The non-residents' commitment against the raise in tuition and fees was evidenced in infOrmal interviews, campus and local newspaper editorials, and a petition by representatives of students and faculty to the University's Board of Regents requesting that they use their powers to influence a reduction of non-resident tuition and fees. In both the Registrar's Office and the veterans'.Affairs Office there was a major increase of non-resident students requesting that 30 their records be sent to prospective new schools of enrollment, and numerous appeals were made to the Financial Aids Office for assistance in relieving the high increase of cost to the student. The position of non-resident students was further demonstrated at a later time by a 43 percent decrease in.non-resident freshmen seeking admission to the Wisconsin State University System for the 1970-71 school year.1 Students were aware that resident tuition and fees had increased about 32 percent in the State University System while non— resident tuition and fees had increased 84 percent.2 The Executive Director of the Wisconsin State University System discussing non- resident enrollment ceilings, indicated they were useless because of the declining non-resident enrollment. He stated, "Tuition and fees have taken care of that. . ."3 It was acknowledged that the non- residents felt strongly about the increased costs and the reasons they believed had brought about the increase. From groups designated with high prior commitment and low prior commitment, a total of 145 subjects were ultimately used in the experiment and were distributed throughout the experimental design as illustrated in Table l. 1"Fewer NOn-Residents Apply to State U's," Pe tomist (Wisconsin State University-Superior), April 29, 1970, p. 3. 2"Increasing Fees Affect Wis. College Enrollments," Peptomist (Wisconsin State university-Superior), NOvember 19, 1970, p. 0 3Jeff Smoller, "High Fees Cut Numbers of NOn-Resident Students," Capital Times (Madison, Wisconsin), September 28, 1970. 31 TABLE 1 THE POPULATION USED IN THE INVESTIGATION Low Prior Commitment High Prior Commitment Low High LOW' High Commitment Commitment Commitment Commitment Low Justification 25 16 ll 6 High Justification 22 30 10 9 Control 8 4 2 2 II. The Experimental Design and Methodology The subjects, students at Wisconsin State University-Superior, were part of an experiment that had 2 x 3 x 2 factorial design, as illustrated in Table l. The conditions of the experiment consisted of subjects designated as having high prior commitment who were placed in a situation where it was necessary to perform a belief discrepant behavior. These subjects were given one of two levels of verbal justification for engaging in the discrepant behavior under two levels of commitment. Commitment was differentiated by the degree of supposed public exposure the subject was to receive while engaging in the belief discrepant behavior. Shmilar groups designated as having low prior commitment were structured as part of the experiment. In addition, control groups having either low or high prior commitment under two levels of commitment but with only one intermediate level of verbal justification were created. The control groups, like the experimental groups, received both a pre-test and a post-test, but the control groups did not perform an interim belief discrepant behavior. 32 In the spring of the 1969-70 school year the instrument, a semantic differential, was administered in a pre-test situation. The pre-test was administered by the regular instructor in previously selected classes. The pre-test, experiment, and post-test had been discussed at length with each participating instructor. At the time of the pre—test the instructor announced: One of the graduate students is compiling statistical data on the basis of the semantic differential. He has asked me if you would be willing to fill out these forms to assist him in the statistical project. YOur name is not needed, as this information will be used only for a statistical exercise. HOwever, it would be appreciated if you would put your social security number in the space provided for your name. These statistics will be analyzed on the computer, which has a program based on social security numbers. This should only take you a short length of time and will be of some assistance to the graduate student. (See Appendix B.) The pre-test semantic differential was then administered to the students who later participated in the experiment.1 ‘Within two weeks of the pre-test, the classroom instructor indicated to the class that they had been chosen to participate in an experiment dealing with the ability of future members of professional groups to develop and to support logical arguments. Specific directions were read to each experimental group by the instructor.2 1The semantic differentials used for both the prestest and post-test are presented in Appendix.A. 2The directions given to each of the experimental groups are presented in Appendix B. 33 The experimental groups were asked to prepare a three-minute Speech supporting high nonrresident tuition for state universities. Prior to starting the outline of their speech, subjects in a high commitment situation were told in part: I will give you twenty munutes to outline your speech and the arguments pertinent to your speech. Please do it on this paper and sign your name in the upper right-hand corner. Prior to starting your writing we would like to focus the camera on each of you, so that we may test the television equipment. Please look at the monitor. At a later date you will be expected to record the speeches on video-tape at the television studio. For the purpose of the demonstration it will be necessary that you identify yourself'and.make your speech as fOrceful as possible. Those people who will see your Speech will assume you believe what you say. (See Appendix B.) In contrast, subjects in low commitment situations were told: I will give yOu twenty minutes to outline your speech and the arguments pertinent to your speech. Please do it on this paper and sign your name in the upper right-hand corner. The student (the word "those" was substituted for the words "the student" in the high justification group) making this study will randomly select most of you to go up to the radio station to record your speech. Please do not be concerned about signing your name. Once the random selections have been made, you will no longer be iden- tified as an individual. The name is merely to identify you so that you may be requested to come and record your speech. Your 34 speech will then be further edited and spliced, so that any possibility of your being identified as an individual is non- existent. (See Appendix B.) The statements read to the subjects also contained a description of what they were being asked to do and how they were expected to com- plete the tasks. In the experimental groups the introductory state- ments were varied slightly depending on the commitment level and justification level of the group. Similar introductory remarks were given to the control group. Near the completion of the directions each group was informed, "You do not have to participate if you choose not to do so." The experimental groups who were asked to perform a belief discrepant behavior were given varied amounts of verbal justifica- tion for committing the act. In high justification conditions, subjects were told that the experiment was being conducted by one of the staff members in conjunction with some of his students under the auspices of awell-known organization. Subjects were told that the experimenter hOped to make both the televised and the audio recordings into a demonstration model to be used in classes throughout the country for problem solving and other logic oriented projects. It was further explained that the experiment could produce new information about cognitive processes. The classroom instructor said: If the group doing the project is able to produce the results they expect, your time will be well Spent because your efforts will lead to new knowledge in cognitive structure, as well as to the deveIOpment of a major learning unit. (See Appendix B.) 35 The subjects in low justification groups were told that the experiment was being conducted by a graduate student who needed their assistance. After the experiment was completed, the classroom instructor of each group concluded by stating: I have another favor to ask of all of you. You have just taken time in preparation for engaging in a publicly televised speech (anonymously recorded radio speech). Previously, I had asked you to help one of our graduate students in a statis- tical project. Unfortunately, after the data was placed on the computer, the tapes were erased and all data was lost. He has asked us to fill out the semantic differential again. As we have already lost some of our class time, we might as well assist the fellow in his project. HOwever, this time if he has problems with losing his data, we will ask him to collect it in another class. (See Appendix B.) The procedures utilized with the experimental groups were duplicated where possible with the control group, except the control subjects were not requested to perform an interim belief discrepant behavior. The subjects in the control situations outlined speeches on the same issue, high tuition and fees for non—resident students in ‘Wisconsin institutions of higher education, but in agreement with their own beliefs. III. Description of the Measuring Instrument Das reports the development of the semantic differential had its beginning in the research of Stagner and Osgood on social 36 stereotypes.1 In 1942, Karwoski, Odert, and Osgood began using the concept of a pair of polar opposites to describe two or more dimen- sions of experience.2 By 1952, Osgood had described the semantic differential and the logic underlying the instrument. In 1957, a definitive description of the instrument, as well as the theoretical and empirical evidence related to it, was published in The Measurement 3 cqubaning. The logic of the instrument was summarized by Osgood and Sebeok, who indicated: 1) The process of description or judgment can be conceived as the allocation of a concept to an experiential continuum defined by a pair of polar terms. 2) JMany different continua of judgment are essentially equivalent and hence may be represented by a single dimension. 3) .A limited number of such continua, representative of the dimensionality of meaningful judgments, can be used to define a semantic space within.which the meaning of any concept can be specified. The semantic differential is in part a combination of controlled association and scaling procedures. The subject is pro- vided with a concept to be differentiated and a set of bipolar 1RheaA. Das, "Semantic Differential Measurement of the Bengali Meaning System," Journal of Psyphological Researchers V61. 13, No. 3, 1969, pp. 131-148. 2Donavan E. Coleman, "The Classical Conditioning of Attitudes TOward Selected Educational Concepts," (unpublished Ed. D. Dissertation, university of South Dakota, vermdllion, South Dakota, 1966), p. 66. 3Charles E. Osgood, George J. Suci, and Percy Tannenbaum, The Measurement of Meaning (Urbana: University of Illinois Press, 1957):- 4Charles E. Osgood and Thomas A, Sebeok (eds.) Ps cho- lin istics: .A Survey of Theory and Research Problems IBIoomington: Indiana.University Press, 1965), pp. 1771178. 37 adjective scales, against which he will make the differentiation.1 In this investigation a seven-step scale was provided for the subject to indicate the direction of his association with the concept and the intensity of that association. Osgood developed two basic forms of the semantic differential. In one form the concept is placed parallel with the adjective pairs. In the other form only one concept appears on each page, and all judgments related to that concept are made consecutively. .Although he felt the first form was not as adequate as the form illustrated below, Osgood's research did not show a significant difference between the 2 results of the two forms. HIGH NON-RESIDENT TUITION good : : : : : : bad small : : : : : : large hot : : : : : : cold The form.illustrated above was utilized in this study. The research of Osgood and his associates yielded three dominant factors or dimensions of semantic space--the evaluative, potency, and activity factors. Adjective pairs representing all three dimensions were used in this investigation. Diab explained: . . . it is evident that the semantic differential procedure does give more information about the meaning of the attitude object than can be obtained from the evaluative dimension alone, and consequently, helps to improve prediction of the Individual's behavior in situations involving the attitude object. 1Osgood, Suci, and Tannenbaum, pp, p13,, p. 20. 21bid., p. 82. 3Lufty N. Diab, ”Measurement of Social Attitudes: Problems and Prospects," in Carolyn W. Sherif and Muzafer Sherif (eds.) Attitude, Ego-Involvement, and Changp, pp, p33,, p. 155. 38 The adjective pairs chosen for this investigation were taken from those originally studied by Osgood and his associates. The pairs chosen met the criteria established by the authors and crossed the three major dimensions of semantic space. The directions for and the semantic differential used in this investigation appear in Appendix A. The reliability of the instrument could be described as the consistency of the instrument. If the instrument is reliable, the measurement units will be similar from time to time. Osgood's study of the test-retest reliability of the subjects on 100 subjects yielded a coefficient of reliability of .85.1 Osgood and his associates men- tion numerous studies that substantiate, to their satisfaction, the reliability of the semantic differential. The validity of the instrument, that is, the question of whether the instrument measures what it is supposed to measure, has undergone extensive analysis. Osgood established the "face validity" of the instrument in The Measurement of Meaning. The criterion of external correlation as presented by Osgood is discussed by Coleman. According to Coleman, Osgood has indicated that on the evaluative factor correlations of .90 or better were found between Thurstone's Attitude Scale and the semantic differential. Similarly, highly significant rank order coefficients with the Guttman scale were Obtained.2 An investigation by Smith concerning the validity of the semantic differential as a means for rating speaker effectiveness led to con- clusions supportive of the validity of the instrument. Smith found 1Osgood, Suci, and Tannenba'um, pp. 311., pp. 126-127. 2Coleman, pp, p13,, pp. 74-75. 39 significant relationships between the semantic differential and student speaker self-ratings, as well as instructor ratings of most- able, least-able subgroups. These and other findings led him to conclude that the semantic differential is a valid means of measuring speaking ability through student self-ratings and is a valid instrument for instructors to use in measuring speaking ability.1 The research of Kelly and Levy led them to conclude, in part, that their results supported the validity of the semantic differential as an adequate way of measuring the connotative meaning of concept.2 Further evidence was presented by Grigg. He indicated that a result favorable to the validity of the semantic differential was illustrated by a group of normal subjects who indicated a significantly greater distance between "ideal self" and neurotic than between neurotic and self.3 As reflected in Table 2, the data collected as part of this investigation lend support to the work of Osgood and his associates. TABLE 2 WITHIN CELLS CORRELATION MATRIX LISTED BY CORRESPONDING SEMANTIC DIFFERENTIAL SCALES EVL - T2 POT - T2 ACT - T2 EVL - F2 0.914266 POT - F2 0.663011 ACT - F2 0.611503 EVL - Tl 0.834311 POT - Tl 0.592291 ACT - T1 0.443542 EVL - F1 0.821319 POT - F1 0.517830 ACT - F1 0.465337 1Raymond G. Smith, "Validation of a Semantic Differential," Sppech MOnographs,.March, 1963, 30: 50-55. 2JaneA. Kelly and Leon H. Levy, "The Discriminability of Concepts Differentiated by Means of the Semantic Differential," Educational and Psychological Measurement, Spring, 1961, 21: 53-57. 3Austin E. Grigg, ”A Validity Study of the Semantic Differential Technique," Journal of Clinical Psychology, April, 1959, 15: 179-181. 40 a. the number 1 after the letter T or F represents pre-test b. the number 2 after the letter T or F represents post-test c. the letter T represents the concept of high non-resident tuition d. the letter F represents the concept of increased fees for non-resident students e. the letters EVL represent the evaluative scale of the semantic differential f. the letters POT represent the potency scale of the semantic differential g. the letters ACT represent the activity scale of the semantic differential Relatively high correlations are Obtained.when one evaluative scale is compared with another evaluative factor or potency is compared with potency and activity is compared with activity scales. In contrast, the correlation between the evaluative scale on the concept related to tuition in the post-test situation and the activity scale on the concept related to fees in the pre-test situation is 0.052760. Although not shown in Table 2, simdlar low correlations existed between all contrasting scales. Such results are expected on the basis of Osgood and his associates' earlier work, and Table 2 is illustrative of the reliability and validity of their work. Osgood's work is further substantiated by the data in Table 3. TABLE 3 STANDARDIZED REGRESSION COEFFICIENTS--INDEPENDENT BY DEPENDENT VARIABLES EVL-T2 POT-T2 ACT-T2 EVL-F2 POT-F2 ACT-F2 EVL - T1 0.510086 0.121833‘ -0.038848 0.212767 -0.053291 0.265830 POT - Tl -0.069975 0.431534 0.034183 -0.l78488 0.117563 0.092861 ACT - Tl 0.124334 -0.096352 0.248890 0.151188 -0.138476 0.263799 EVL - F1 0.345777 -0.211533 0.063478 0.567140 0.003201 -0.227359 POT - F1 0.074289 0.247899 0.001317 0.146807 0.376258 0.110554 ACT - F1 -0.118803 0.100530 0.306915 -0.145531 0.193619 0.220053 41 a. the number 1 after the letter T or F represents pre-test b. the number 2 after the letter T or F represents post-test c. the letter T represents the concept of high non-residenttuition d. the letter F represents the concept of increased fees for non-resident students e. the letters EVL represent the evaluative scale of the semantic differential f. the letters POT represent the potency scale of the semantic differential g. the letters ACI‘ represent the activity scale of the semantic differential The relative value per unit of evaluation is higher when corresponding scales are related; i.e. , the standardized coefficient is higher between evaluative (EVL) scale and evaluative scale than between‘ evaluative scale and activity (ACI‘) scale or potency (POT) scale. It is expected from the writings of Osgood that such relationships would exist, and in effect this data lends support to validity and reliability studies of the semantic differential previously conducted by Osgood and others . IV. Hypotheses Used in this Investigation The semantic differential can be uSed as a generalized attitude measurement technique , provided the evaluative adjective pairs are used. 1 The instrument has had extensive use and is a generally accepted instrument for measuring attitude and attitude change. In this investigation a semantic differential instrument was used as a pre-test and post-test measure in relation to the following hypotheses: 1Fred N. Kerlinger, Foundations of Behavioral Research, (New York: Holt, Rinehart, ancfiIIinston, Inc., 1964), p. 579. 42 H1: .A change of attitude can be induced and explained by dissonance in subjects who have a high degree of prior commitment. H2: When subjects are engaging in a belief discrepant behavior, a high commitment situation and a low commitment situation do not result in the same amount of attitude change. H3: Subjects receiving verbal justification for engaging in a belief discrepant behavior do not eXperience the same amount of attitude change as other subjects receiving less justification for engaging in a belief discrepant behavior. These hypotheses were evaluated in light of data analyzed through an analysis of covariance. V. Statistical Techniques Used for the Analysis of the Data The design of this study might best be described by what Campbell and Stanley call "the nonequivalent control group design".1 The subjects in this investigation were chosen from regularly scheduled college classes. With the exception of the control group, students were not assigned at random to treatment groups. The experiment was con- ducted within the confines of the regular classroom; the pre-tests and post-tests were also administered in the classroom. If the appropriate assumptions are met, the design should control the main effects of history, maturation, instrumentation, and testing.2 1Donald T. Campbell and Julian C. Stanley, Experimental and ansi-Experimental Designs for Research, (Chicago: Rand'MCnally and COmpany, 1963), p. 47. 21bid., p. 48. 43 Due to the nature of the experimental design, a multivariate analysis of covariance was chosen to analyze the data. The analysis of covariance is explained by Kerlinger: . .a form of analysis of variance that tests the significance of the differences between.means of final experimental data by taking into account and adjusting initial differences in the data. That is, the analysis of covariance analyzes the differences between experimental groups on Y after taking into account either initial differences in the Y measurei or differences in some pertinent independent variable. In this investigation the pre-test scores were used as the covariates. This form of analysis will help to control the group differences statistically. The analysis should result in a smaller estimate of experimental error in comparison to an analysis of variance of the measures.2 The net result of using the analysis of covariance should result in a more precise analysis of the data. Campbell and Stanley warn that in designs similar to the one used in this investigation a problem with regression is possible.3 The purported regression is often attributed to either correlated measures or the selection of subjects on the basis of extreme pre-test scores. It is expected the more deviant the score, the larger will be the error of measurement.4 .As this investigation involves subjects lKerlinger, 9p. c_i_p., p. 347-348. zAllen L. Edwards, Experimental Design in Psychological Research (New YOrk: HOlt, Rinehart, and Winston, Inc., 1968), p. 3Campbell and Stanley, pp. _c_2_i;t_., p. 47-50. 4Donald T. Campbell and Julian C. Stanley, "EXperimental and Quasi-Experimental Designs for Research on Teaching," in N. L. Gage (ed.) Handbook of Research on Teaching (Chicago: Rand.McNally and Company, 1963), p. 181. 44 with extreme stands and a potential for correlated measures, a multivariate analysis of covariance was chosen to analyze the data. The analysis was done with a Fortran program written by Finn of the State University of New York at Buffalo and modified for use at Michigan State University by Wright.1 The computer analysis chosen for the investigation allowed not only for a univariate or multi- variate analysis of covariance but also for related transformations, cell means, least square estimates, regression analysis, and canonical correlations, as well as hypotheses tests related to the investigation2 VI. Summary The 145 subjects in this investigation were students at Wisconsin State University-Superior. The University is a member of a nine-university state system which had to implement a dramatic increase in non-resident tuition just prior to the beginning of the 1969-70 school year. Circumstances allowed for almost no advance notice of the high increase in fees to either new or returning non- resident students. Evidence led to the conclusion that all students 3 had strong feelings about this issue. » Arbitrarily, resident stu- dents were designated as having low prior commitment and non-residents 1An abbreviated description of the forms of analysis and program created by Jeremy Finn and modified by David Wright appears in Appendix C. 2An abbreviated description of the forms of analysis and program created by Jeremy Finn and modified by David Wright appears in Appendix C. 3The Coordinating Council of Higher Education in Wisconsin has reported that tuition and fees in the Wisconsin State University System rose 91.1 percent between 1967 and 1970 with a corresponding decrease of 41.5 percent in new freshman enrollment during the same period. 45 were designated as having high prior commitment. .Although it was assumed that most of the potential subjects had strong feelings about the issue, it was determined for the purpose of this investigation to designate the resident and non-residents as separate groups. All the subjects designated as having either low prior commitment or high prior commitment were either placed in experimental groups where they were forced to participate in a belief discrepant behavior or in a control group. The experimental and control group subjects were placed in a high commitment Situation (prOposed- televised-publicly identified use of statements) or in a low commitment Situation (prOposed-anonymous-non video recorded use of statements). The subjects in the experimental groups were given either a high or low amount of verbal justification for participating in the belief discrepant behavior. subjects in the control situations were given one level of justification. The investigation had a 2 x 3 x 2 factorial design. A semantic differential was administered in a pre-test and post-test Situation. The resulting data fit into one of twelve cells and was subjected to multivariate analysis of covariance. The hypotheses to be tested were related to the ability of dissonance to induce and to explain attitude change in individuals, as well as to explore the effect of commitment and justification on the proposed attitude change. CHAPTER IV ANALYSIS OF DATA This investigation was centered upon the ability of dissonance to induce and to explain a change of attitude in subjects with a high degree of prior commitment about an issue. The effects of commitment and justification were also reviewed in the analysis to determine if they contributed to any purported change of attitude. The data consisted of semantic differential scores which were analyzed through the use of both multivariate and univariate analyses Of covariance. The analysis of the data was divided into two primary areas: 1) the applicability and apprOpriativeness of the measuring instrument and statistical techniques used in the investigation, and 2) a series of statistical tests related to the hypotheses of the investigation. Specifically tested were the effects of: 1. low prior commitment y_s_ high prior commitment 2. low justification Y2 control and high justification _v_s_ control 3 . low commitment yp high commitment 4. prior commitment 5 justification interaction 5. prior commitment x_ commitment interaction 6. justification )__(_ commitment interaction 7. prior commitment 1‘. justification x_ commitment interaction 46 47 I. ApprOpriativeness of the Instrument and Analysis The independence of the scale values of the semantic differential is illustrated in Table 4. TABLE 4 MATRIX OF CORRELATIONS.AFTER ADJUSTING FOR THE EFFECTS OF THE CO-VARIABLES WITH RELATED VARIANCE AND STANDARD DEVIATION EVL - T2 POT - T2 ACT - T2 EVL - F2 POT - F2 ACT - F2 EVL - T2 -------- POT - T2 -0.l63244 -------- ACT - T2 0.029946 0.397267 -------- EVL - F2 0.757459 -0.096833 0.020242 -------- POT - F2 -0.1l3022 0.546281 0.329791 -0.066424 -------- ACT - F2 0.121057 0.069312 0.478542 0.091115 0.216737 -------- Degrees I Standard of Variable variance Deviation Freedom 127 EVl - T2 369.615842 19.2254 POT - T2 34.137523 5.8427 ACT - T2 16.965510 4.1189 EVL - F2 471.851282 21.7221 POT - F2 42.257394 6.5006 ACT - F2 17.702325 4.2074 the letter T represents the concept of high non-resident tuition the letter F represents the concept of increased fees for non- resident students the letters EVL represent the evaluative scale of the semantic differential the letters POT represent the potency scale of the semantic differential the letters ACT represent the activity scale of the semantic differential 48 The independence of the scale values of the semantic differential is illustrated by the low correlations between scales. The correlation between the evaluative scale on the tuition concept at the post-test level and activity scale on the tuition concept at the post-test level is 0.029946 and is exemplary of the independence between scales. This fact is further substantiated by the high correlation between concepts on the same scale, for example, a correlation of 0.757459 between the post-tests on the evaluative scale for the two concepts related to tuition and fees. A chi square test of hypothesis of no association between dependent and independent variables showed that a covariate analysis was needed. The high degree of relationship between pre-test and post-test variables is shown in Table 5. TABLE 5 REGRESSION ANALYSIS WITH THE SIX PRE-TESTS AS COVARIATES Variable Multiple R F P Less Than EVL - T2 0.8564 58.2507 0.0000 POT - T2 0.6458 15.1403 0.0001 ACT - T2 0.5129 7.2570 0.0001 EVL - F2 0.8165 42.3427 0.0000 POT - F2 0.5286 8.2091 0.0001 ACT - F2 0.5489 9.1287 0.0001 df 6 for hypothesis 127 for error 49 a. the letter T represents the concept of high non-resident tuition b. the letter F represents the concept of increased fees for non- resident students c. the letters EVL represent the evaluative scale of the semantic differential d. the letters POT represent the potency scale of the semantic differential e. the letters ACT represent the activity scale of the semantic differential The ability to predict post-test results from the pre-test is indicated by the high correlations. The degree of relationship is Significant at p=.01 level on all scales and all concepts. Das has explained the usefulness of canonical analysis in multi- variate investigations. She stated: "Canonical analysis deals with the correlations within and between two sets of variables, and gives the linear 1 The combinations of the variables with maximum.correlation in each set." canonical correlations and the percent of variation in the dependent variables that the correlations accounted for are illustrated in Table 6. TABLE 6 CANONICAL CORRELATIONS AND THE PERCENT OF VARIATION IN DEPENDENT VARIABLES Canonical Correlation Percent of variation 1 0.8616 12.3722 2 0.6778 7.6566 3 0.4925 4.0424 4 0.4002 2.6696 5 0.1856 0.5741 6 0.1203 0.2414 Total % = 27.5563 thea S. Das, "An Application of Factor and Canonical Analysis to Moltivariate Data," The British Journal of Mathematical and Statistical Ppychology, May, 1965, V61. 18, Part I, p. 59. 50 The first canonical correlation (.8616) represents the degree of relationship between the weighted scores of the pre-test (six variables) and the post-test (six variables) when the weights have been chosen to get maximm correlation between the two matrices. After analysis of the canonical correlation, 72.4437 percent of the variance in the dependent variables remained unexplained. Due to the design of the study, the unexplained variance would not be attributed to subject differences. II. Tests of Significance The first hypothesis to be reviewed in this investigation was related to an.attempt to explore if a change of attitude could be induced and explained by dissonance in subjects who had a high degree of prior commitment. Post-test scores with the pre-test as the covariate were subjected to an analysis of covariance. The cell means for all variables are presented in Table 7. In order to determine if a statistically significant change of attitude had occurred in the experimental groups , a multivariate analysis of covariance was utilized. A summary table of the multi- variate analysis of covariance is presented in Table 8. As indicated in Table 8, no statistically significant differences were found between the equality of mean vectors. On this basis, the null hypothesis related to a change of attitude induced and explained by dissonance in subjects with a high degree of prior commitment could not be rejected. 51 TABLE 7 CELL MEANS FOR BOTH POST-TEST AND PRE-TEST RESULTS Low High Justification Justification Control Low High Low High Low High Cormitment Commitment Commitment Commitment Commi tment Commitment EVL - T2 99.68 97.50 76.05 87.07 110.00 104.25 POT - T2 46.32 46.31 47.27 44.20 42.75 40.00 ACT - T2 25.76 26.94 25.23 24.13 25.13 23.50 EVL - F2 99.00 93.63 79.50 86.70 101.63 104.00 POT - F2 43.32 44.75 45.91 42.87 44.88 39.25 Low .ACT - F2 27.04 26.06 24.05 24.10 23.88 23.00 rior EVL - Tl 94.24 110.25 78.68 91.77 116.00 109.25 POT - T1 47.40 46.31 51.36 45.20 45.63 42.50 ACT - T1 25.80 28.13 24.91 25.13 26.13 24.00 EVL - Fl 97.32 97.94 76.36 89.23 117.00 104.75 POT - Fl 44.16 47.13 46.73 43.33 41.13 41.00 ACT - Fl 26.44 27.88 26.59 25.73 24.75 24.50 EVL - T2 83.73 74.33 56.80 88.89 133.00 96.50 POT - T2 47.09 42.83 54,40 44.11 37.00 41.00 ACT - T2 27.73 25.67 29.20 24.22 24.00 23.50 EVL - F2 80.45 81.17 52.10 92.22 131.50 88.00 POT - F2 47.00 40.00 56.00 40.44 37.00 41.50 High ACT - F2 27.91 25.17 28.90 26.44 24.00 25.50 Prior EVL - Tl 81.73 83.00 63.80 91.67 126.50 110.50 POT - Tl 49.73 41.67 51.60 47.89 35.50 47.00 ACT - Tl 28.45 23.67 27.30 25.56 24.00 23.50 EVL - Fl 77.91 78.00 61.90 89.67 136.00 110.00 POT - Fl 46.73 40.67 47.70 43.44 43.00 45.00 ACT - F1 28.36 23.33 26.40 26.78 24.00 25.00 a. the number 1 after the letter T or F represents pre-test b. the number 2 after the letter T or F represents post-test c. the letter T represents the concept of high non-resident tuition d. the letter F represents the concept of increased fees fOr non-resident students e. the letters EVL represent the evaluative scale of the semantic differential f. the letters POT represent the potency scale of the semantic differential g. the letters ACT represent the activity scale of the semantic differential 52 TABLE 8 SUMMARY TABLE OF P RATIOS--LISTED BY EXPERIMENTAL CONDITION FOR.MULTIVARIATE TESTS OF EQUALITY OF MEAN VECTORS P Less Condition df F Than Low Prior Commitment y§_High Prior 6 and 122 .1486 0.3385 Commitment Low Justification y§_Control and High 12 and 244 .7577 0.6937 Justification y§_Control Low Commitment yp High Commitment 6 and 122 .4147 0.2143 Prior Commitment §_Justification 12 and 244 .7143 0.7371 Interaction Prior Commitment §_Commdtment 6 and 122 .0446 0.0648 Interaction Justification §_Commitment Interaction 12 and 244 .0541 0.4001 Prior Commitment x Justification x_ 12 and 244 .2069 0.2787 Commitment Intéraction Although the multivariate analysis of covariance did not indi— cate any statistically significant data, a series of univariate analyses were performed. The univariate analysis was chosen due to the indepen- dence of the scales of the semantic differential. The univariate analysis would allow for a review of the data related to each of the three scales of the semantic differential and would improve the investi- gation. Table 9 presents a summary of the univariate analysis of low prior commitment v_s_ high prior commitment. groups and the high prior commitment groups were Significantly The low prior commitment different statistically only on the activity scale, and then only on the concept related to fees. SUI/MARY TABLE OF ANALYSIS OF LOW PRIOR CO'VMITIIIENI‘ V_S_ HIGH PRIOR COMVIITMENI‘ ( 1 df for Hypothesis) (127 df for Error ) Variable Between P Post-Test Mean Univariate Less Measure Square F Than EVL - T2 56.5198 0.1529 0.6965 n.s. POT - T2 20.7510 0.6079 0.4371 n.s. ACT - T2 41.5512 2.4492 0.1201 n.s. EVL - F2 168.4781 0.3571 0.5513 n.s. POT - F2 83.6329 1.9791 0.1620 n.s. ACT - F2 101.5838 5.7384 0.0181 * a = Significant (P=.05) NOn-Significant n.s. The Effects of Commitment: The second hypothesis of this investigation was related to determining if, while engaging in an act which is discrepant with one's private belief, a high commitment situation (prOposed-televised- publicly identified use of statements) or a low commitment situation (proposed-anonymouS-non video recorded use of statements) results in the subjects having a greater change of attitude. .A univariate analysis with the six covariates eliminated was in part utilized to examine the effect of commitment. Table 10 illustrates the effect of the commit- ment variable. The only statistically Significant difference in Table 10 is related to the potency scale of the semantic differential, and then only on the concept related to fees. The statistical information pertinent to the second hypothesis did not allow for a rejection of the related null hypothesis. 54 TABLE 10 SUMMARY TABLE OF ANALYSIS OF LOW COMMITMENT V§_HIGH COMMITMENT ( 1 df for Hypothesis) (127 df for Error ) variable Between P Post-Test Mean Univariate Less Measure Square F Than EVL - T2 871.9802 2.3592 0.1271 n.s. POT - T2 30.0520 0.8803 0.3499 n.s. ACT - T2 23.1285 1.3633 0.2452 n.s. EVL - F2 553.7841 1.1736 0.2808 n.s. POT - F2 226.7379 5.3656 0.0222 * ACT - F2 14.1801 0.8010 0.3725 n.s. * = Significant (P=.05) NOn-Significant i" (D II The Effects of Justification: The third hypothesis of this investigation was related to determining if the degree of verbal justification given subjects for engaging in a belief discrepant behavior had any effect on a prOposed change of attitude. A.summary table of the univariate analysis of low justification y§_control and high justification.y§_control is presented in Table 11. No statistically Significant differences are found in Table 11, and the null hypothesis related to justification could not be rejected. Interaction Effects: Interaction effects were also assessed through univariate analysis. Summary tables of interaction effects are presented in Tables 12, 13, 14, and 15. SUMVIARY TABLE OF ANALYSIS OF LOW JUSTIFICATION VS CONTROL AND HIGH JUSTIFICATION V§_CONTROL ( 2 df for Hypothesis) (127 df for Error ) Variable Between P Post-Test Mean Univariate Less Measure Square F Than EVL - T2 154.5018 0.4180 0.6593 n.s. POT - T2 22.1179 0.6479 0.5249 n.s. ACT - T2 8.9111 0.5252 0.5927 n.s. EVL - F2 453.6161 0.9614 0.3852 n.s. POT - F2 14.9056 0.3527 0.7035 n.s. ACT - F2 21.5912 1.2197 0.2988 n.s. * = Significant (P=.05) n.s. = NOn-Significant TABLE 12 SUMWARY TABLE OF ANALYSIS OF PRIOR CONMITMENI‘ X_JUSTIFICATION INTERACTION ( 2 df for Hypothesis) (127 df for Error ) Variable Between P Post-Test Mean Univariate Less Measure Square F Than EVL - T2 17.2530 0.0467 0.9545 n.s. POT - T2 52.0564 1.5249 0.2216 n.s. ACT - T2 4.5402 0.2676 0.7657 n.s. EVL - F2 88.7612 0.1881 0.8288 n.s. POT - F2 82.7875 1.9591 0.1453 n.s. ACT - F2 19.3406 1.0925 0.3385 n.s. 8’ ll :3 U) ll Significant (P=.05) NOn-Significant 56 TABLE 13 SUMMARY TABLE OF ANALYSIS OF PRIOR COIVMITMENT _X COD/MITMENT INTERACTION ( 1 df for Hypothesis) (127 df for Error ) Variable Between P Post-Test Mean Univariate Less Measure Square F Than EVL - T2 147.5498 0.3992 0.5287 n.s. POT - T2 80.4979 2.3580 0.1272 n.s. ACT - T2 16.9448 0.9988 0.3196 n.s. EVL - F2 1452.3608 3.0780 0.0818 n.s. POT - F2 330.1185 7.8121 0.0060 * ACT - F2 0.2159 0.0122 0.9123 n.s. * 5 Significant (P=.05) n.s. = Non-Significant TABLE 14 SUI/MARY TABLE OF ANALYSIS OF JUSTIFICATION X;CThhHfllflflvI INTERACTION’ ( 2 df for Hypothesis) (127 df for Error ) variable Between P Post-Test Mean univariate Less Measure Square F Than EVL - T2 646.8034 1.7499 0.1780 n.s. POT - T2 8.2731 0.2423 0.7852 n.s. ACT - T2 18.8797 1.1128 0.3319 n.s. EVL - F2 514.1384 1.0896 0.3395 n.s. POT - F2 52.1476 1.2340 0.2946 n.s. ACT - F2 12.9634 0.7323 0.4829 n.s. * = Significant (P=.05) n.s. = NOn-Significant 57 TABLE 15 SUNMARY TABLE OF ANALYSIS OF PRIOR COWITMENI‘ X JUSTIFICATION X C(MVIITMENI INTERACTION ( 2 df-for Hypothesis) (127 df for Error ) Variable Between P Post-Test Mean Univariate Less Measure Square F Than EVL - T2 236.0958 0.6388 0.5297 n.s. POT - T2 76.8182 2.2503 0.1096 n.s. ACT - T2 17.0351 1.0041 0.3693 n.s. EVL - F2 799.3436 1.6941 0.1879 n.s. POT - F2 153.0129 3.6210 0.0296 * ACT - F2 25.9137 1.4639 0.2353 n.s. Significant (P=.05) Non-Significant The only statistically significant interaction effects were found twice on the potency scale, but only on the concept related to fees. The two statistically Significant interaction effects were related to prior commitment §_justification interaction and prior commitment §_commitment interaction. 111. Discussion The statistical results of this investigation failed to support any of the three hypotheses of the investigation. The first hypothesis predicted that a change in attitude could be induced and be explained by dissonance in subjects who have a high degree of prior commitment about an issue. The failure of the data to support the hypothesis could be explained by assimilation-contrast theory. Assimilation-contrast theorists (social judgment-involvement approach theorists) predict that the more involved the subject is in his stand, the less likely he is to 58 change his attitudes. Those using the social judgment-involvement approach expect that a great discrepancy between one's own position and that advocated in a contrary communication or role will result in little or no attitude change. In a Speculative manner it is possible to explain the results of the data in terms of dissonance theory. As dissonance reduction can be accomplished by means other than a change of attitude, other avenues of dissonance reduction must be reviewed. On the basis of the analysis of the outlines of the speeches prepared by subjects from both low prior commitment and high prior commitment groups in a high justification and low commitment situation, another form of dissonance reduction is plausible. Subjects in the forced compliance situation reduced dissonance by creating arguments that were really in line with their own personal beliefs but yet fulfilled the requirements of the forced compliance situation. In.brief, arguments made were in support of high non-resident tuition and fees but were stated in such a manner that little or no dissonance would be created for the individual Opposing high non-resident tuition and fees. The following quotes from two subjects in the high justification and low commitment situation are exemplary of such a proposed.mode of dissonance reduction: High out-of-state tuition will keep a lot of trouble- making Eastern Slobs out of the great Wisconsin educational system. Their ideas and beliefs are too radical for the simple- munded idiots of the Midwest, and by keeping these communists out, the State Universities will be able to do as they wish 'without fear of student unrest. Another example of such an approach is seen in a second subject's outline: 59 I High tuition is good because it puts out of state students in poverty. At It forces one to hold two jobs. B. It cuts down on student consumer spending. 11 High tuition is good because it forces down student enrollment and.makes dorms that could be used to house students shut down. A- It's good because it cuts down funds necessary to run the school B. 1 Other subjects in this experimental group, not using a Similar approach,frequently used arguments that were originally released in the mass media as explanation for the dramatic increase in non-resident tuition and fees. It had been previously established that such argu- ments were not acceptable to the majority of the subjects. The data related to the first hypothesis is supportive of the position taken by social judgment-involvement theorists, but other forms of evidence do not exclude a dissonance interpretation of the results of the investigation. The second hypothesis was to explore if a variation of commit- ment (amount of supposed public exposure) would affect the degree of prOposed attitude change. Although in the univariate analyses there are isolated cases of significantly different data concerning the commitment variable, no main effect or clear interaction effects for commitment were discovered. As a result of the lack of statis- tical Significance, the null hypothesis could not be rejected. 1These quotes are from the original outlines prepared by subjects in one of the experimental groups of this investigation. 60 The third hypothesis predicted that the degree of prOposed attitude change would not be the same for subjects receiving high verbal justification and subjects receiving a lesser amount of verbal justification for engaging in a belief discrepant behavior. The data failed to support a rejection of the third null hypothesis . The failure of the data to support the third hypothesis can be explained in one or more ways. One interpretation of the data would be that neither level of justification was sufficient to justify a Significant change, nor was the justification adequate to arouse dissonance. Another interpretation would be that both levels of justification were too large to arouse dissonance but insufficient to justify a change of attitude from the standpoint of the social judgment-involvement approach. On the basis of the debriefing sessions and selective follow-up interviews, a probable speculation is that subjects felt a tension that might be described as dissonance; but, if in fact they did have dissonance, a means of dissonance reduction other than a change of attitude was used, and the role of justification became insignificant. Unfortunately, no measure of the subjects perception of the magnitude of justification was made, and any interpretation of data related to justification must be speculative. IV. .A Summary This investigation was centered upon the ability of dissonance to induce and to explain a change of attitude in subjects with a high degree of prior commitment about an issue. The effects of commitment and justification.were also studied to determine if they contributed to the prOposed change of attitude. Data gathered from pre-test and post-test 61 semantic differential scores were analyzed with multivariate analysis of covariance and univariate analysis techniques. Significance tests were made on: 1. low prior commitment prhigh prior commitment 2. low justification y§_control and high justification y§_ control 3. low commitment y§_high commitment 4. prior commitment §_justification interaction 5. prior commitment §_commitment interaction 6. justification §_commitment interaction 7. prior commitment x justification x commitment interaction The multivariate analysis of covariance did not yield any statistically significant results. The univariate analysis was justified on the basis of the independence of the varied scales of the semantic differential and produced isolated instances of significantly different data, for example, P = 0.0222 on the potency scale for the post-test on the concept related to fees. On the basis of the statis- tical evidence accumulated, it was not possible to reject any of the three null hypotheses related to either a change of attitude being induced and explained by dissonance in subjects with high prior commit- ment about an issue,or a difference in degree of prOposed attitude change as a result of varied amounts of commitment (prOposed public exposure), or a difference in degree of proposed attitude change as a result of varied levels of verbal justification for engaging in a belief discrepant behavior. The failure of the data to support any of the three alternate hypotheses could be explained from the position of the social judgment- 62 involvement approach theorists. They expect that a great discrepancy between one's own position and that advocated in a contrary communica- tion or role will result in little or no attitude change. In a Speculative manner the results of the investigation could be explained by subjects finding an alternate means other than a change of attitude for dissonance reduction. Examples were given in this chapter of such possible forms of dissonance reduction. Regardless of the acceptance of either of the two explanations for the failure to create and explain a change of attitude by dissonance, the two alternate hypotheses related to commitment and justification could not be proven by the data of this investigation. .A lack of a main effect for justification could be explained by subjects selecting an alternate means of dissonance reduction other than a change of attitude, thus reducing the importance of the justification variable. From the standpoint of the social judgment-involvement approach theorists, the lack of a main effect for justification could be explained by indicating that the amount of justification was insufficient to produce anything of Significance. CHAPTER‘V SUMMARY AND CONCLUSIONS The purpose of this investigation was to detenmine if a change of attitude could be induced and explained by dissonance in subjects with high prior commitment about an issue. The primary questions con- sidered in the study were: 1. Can a change of attitude be induced and explained by dissonance in subjects who have a high degree of prior commitment? When engaging in an act which is discrepant with one's private beliefs, does a high commitmentsituation (proposed-televised-publicly identified use of state- ments) or a low commitment Situation (proposed-anonymous- non video recorded use of statements) result in the subjects having a greater change of attitude? Does more or less verbal justification for subjects engaging in a discrepant behavior result in greater attitude change? A review of related literature was made. Emphasis was placed on the writings of dissonance theorists and theorists using the social judgment-involvement approach to attitude change. The contrasts in the literature related to the two theories made the possible results of the investigation.more speculative. It was clear that neither dissonance theory nor the theory related to the social judgment- 63 64 involvement approach to attitude change could be ignored in the interpretation of the results of this investigation. 1. Procedures of the Study The subjects in this investigation were 145 students at Wisconsin State university-Superior. The University, just prior to the beginning of the 1969-70 school year, had to implement a dramatic increase in non-resident tuition and fees. .Although evidence led to the conclusion that most students had strong feelings about the tuition and fee increase, on an arbitrary basis resident and non-resident stu- dents were respectively designated as having either low prior commit- ment or high prior commitment. .All the subjects designated as having either low prior commit- ment or high prior commitment were assigned either to one of the experimental groups where they were fOrced to participate in a belief discrepant behavior or to control groups which did.not perfOrm.a belief discrepant behavior. The experimental and control group subjects were placed in a high commitment situation (proposed-televised-publicly identified use of the statements) or in a low commitment situation (prOposed-anonymous-non video recorded use of the statements). The subjects in the experimental groups were given one of two levels of verbal justification fOr participating in the belief discrepant behavior. subjects in the control situations were given one level of justification. A semantic differential was administered in a pre-test and post-test situation. A multivariate analysis of covariance and a univariate analysis were used in analyzing the resulting data. The basic design of the investigation was a 2 x 3 x 2 factorial design. 65 II. Summary of Findings Data gathered from pre-test and post-test semantic differential scores were analyzed with multivariate analysis of covariance and univariate analysis techniques. Significance tests were made on: 1. low prior commitment y§_high prior commitment 2. low justification y§_control and high justification yp control 3. low commitment y§_high commitment 4. prior commitment §_justif1cation interaction 5. prior commitment §_commitment interaction 6. justification §_commitment interaction 7. prior commitment §_justification §_commitment interaction The multivariate analysis of covariance did not yield any statistically significant results; however, the univariate analysis produced some statistically significant results, but not in any consistent manner. On the basis of the statistical evidence accumulated, none of the three null hypotheses related to this investigation could be rejected. III. Conclusions The failure of the data to support the first hypothesis of this investigation, which was related to the ability of dissonance to induce and explain a change of attitude in subjects with a high degree of prior commitment, would be conducive to interpretation from the theoretical position of the social judgment-involvement approach to attitude change. The data from this investigation in large supports the findings of Rand in his empirical comparison of the social 66 judgment-involvement approach and dissonance theory.1 A.speculative explanation for the failure of the data to warrant rejection of the first null hypothesis can be found in dissonance theory. As dissonance can be reduced by means other than a change of attitude, it can be suggested that an alternate form.of dissonance reduction was chosen by the subjects. A review of some of the experimental subjects' speech outlines indicates that some subjects did reduce dissonance by slanting their outlines in such a manner as to reduce their dissonant situations. No main effects or clear interaction effects for either commit- ment or justification were found. The data related to the second and third hypotheses of this investigation did not allow for the rejection of either the second or third null hypotheses. The failure to confirm the alternate hypotheses of this study could not be attributed to the measuring instrument. The high correlations of evaluative scales with evaluative scales, potency scales with potency scales, and activity scales with activity scales indicated the reliability of the semantic differential used in this investigation. The independence Of the semantic differential scales was demonstrated by the low correlations between such scales as the evaluative scale and the activity scale. In general the data from this investigation demonstrated the suitability of the semantic differential as a research tool for studies of this type. NO generalizations related to the hypotheses of this investiga- tion appear valid. Although the data does support the position taken lRand, pp. _c_i_t_., p. 112801129-A. 67 by social judgment-involvement theorists, alternative dissonance explanations are plausible and further research is suggested. IV. Implications for Further Research On the basis of the limitations and conclusions of this study, research in the following specific areas is warranted: l. A further investigation with appropriate controls for alternate modes of dissonance reduction to determine if dissonance can induce and can explain attitude change in subjects with.what has been labeled as a high degree of prior commitment in this study. 2. An investigation to determine the nature Of verbal justification as perceived by the experimental subject. 3. A thorough study of the impact of role playing on producing long-term.changes of attitude. LIST OF REFERENCES LIST OF REFERENCES Books / Aronson, Elliot. "lhe Psychology of Insufficient Justification: An Analysis of Some Conflicting Data," in Shel Feldman (ed.). Cognitive Consistency: Motivational Antepedems and Behavioral Consequents. New York: Academic Press,( 1966. l/Brelmm, Jack W. and Arthur R. Cohen. lorations in'Cogpitive Dissonance. New York: John 1 ey and Sons, Inc., 1962. Bruning, James L. and B. L. Kintz. Corputational Handbook of Statistics. Glenview, Illinois: Scott, Foresman, and Company, 1968. Campbell, Donald T. and Julian C. Stanley. Zpli‘t'I‘perimental and giafii Experimental Desi for Research. cago: a y and Company, 196 . Campbell, Donald T. and Julian C. Stanley. "Experimental and Quasi Experimental Designs for Research on Teaching," in N. L. Gage (ed.) . Handbook of Research on Teachirg, Chicago: Rand McNally andCompany, 1963. Diab, Lufty N. "Measurement of Social Attitudes: Problems and Prospects," in Carolyn W. Sherif and Muzafer Sherif (eds.) . Attitude LE o- Involvement; and Change . New York: John Wiley ancISons, nc., 1967. ‘ Edwards, Allen L. Writental Desi? inPsychological Research. New York: 0 t, e art an inston, Inc., 1968. Edwards, Allen L. Statistical Analysis for Students in Psychology and Education. MW York: Rinehart and Company, Inc. , 1946. Feldman, Shel (ed.) . Cognitive Consistency: Motivational Antecedents and Behavioral Consequents. New York: Academic Press, I966. I/ Festinger, Leon. A Theory of Cogpitive Dissonance. Stanford, California: *Stanford university Press, 1957. Gage, N. L. (ed.) . Handbook of Research on Teachipg. Chicago: Rand McNally and Company, 1963. T 68 69 Harper, Robert J. C., Charles C. Anderson, Clifford M. Christensen and Steven M. Hunka. The Cognitive Processes: READINGS. Englewood Cliffs, fiewTersey: Prentice 112111, Inc., I964. Kerlinger, Fred N. Foundations of Behavioral Research. New York: Holt, Rinehart and Winston, Inc., 1964. Kogan, Nathan and Michael A. Wallach. Risk Taking--A Study in Cognition and Personality. New Yorl? Holt, Rinehart and Winston, Inc., 1964. AcGuire, William J. "The Current Status of Cognitive Consistency Theories," in Shel Feldman (ed. ). Cognitive Consistency: Motivational Antecedents Moral Cons puents New York: Academic Pres , 1966. Osgood, Charles E. ,George J. Suci, and Percy H. Tannenbaum. The Measurement of Meanipg. Urbana, Illinois. University—of Illinois Press, 1957. Osgood, Charles E. and Thomas A. Sebeok (eds.). Psycholinguistics: A Survey of Theory and Research Problems. Bloomington, Indiana: Indiana University Press, 1965. I/ Rosenberg, Milton J. "Some Limits of Dissonance: Toward a Differential View of Counter—Attitudinal Performance," in Shel Feldman (ed.). Cognitive Consistency: Motivational Antepedents and Behavioral Consequents . New York: Academic PW w Schramm, Wilbur (ed.). The Science of Human Communication: New Directions and New Findings in CommnmicaTion Research. New York: BasicTooks, Inc., 1963. l/Secord, Paul J. and Carl W. Backman. Social Psycholpgy. New York: McGraw- Hill Book Company, 1964. /Sherif, Carolyn W. and Muzafer Sherif .(edS.) . Attitude L Ego-Involvement, . and Change. New York: John Wiley and Sons, Inc. , 1961. Sherif, Carolyn W. ,Muzafer Sherif, and Roger B. Nebergall. Attitude. and Attitude Change: The Social Judgment- Involvement A pproach.‘ ' delphia: \W. B. Taunders Company, 1965. \/ Sherif, Muzafer and Carl I. Hovland. Social Judgment: Assimilation and Contrast Effects in Communication and AttituEE Change. New Haven: Yale University Press, 1961. Staats, Arthur W. Human Learnings: Studies Extending Conditioning Principles to Complex Behavior. New York: Holt, Rinehart and Winston, Inc., 1964. 70 :,//Whittaker, James O. ”The Resolution of the Communication Discrepancy Issue," in Carolyn W. Sherif and Muzafer Sherif (eds.). .Attitude, Ego-Involvement, and Change. New York: John Wiley and Sons, Inc., 1967. Winer, B. Y. Statistical Principles in Experimental Design. New York: MtGraw—Hill Book Company, 1962. Periodicals kégohen, Arthur R. ”Attitudinal Consequences of Induced Discrepancies Between Cognitions and Behavior," Public Opinion Quarterly, 24:297-318, 1960. Das, Rhea S. "An.Application of Factor and Canonical Analysis to Multivariate Data," British Journal of Mathematical and Statistical Psychology, 18:57467, May, 1965. Das, Rhea S. "Semantic Differential MBasurement of the Bengali Mbaning System," Journal of Psychological Researches, 13:131-148, 1969. Endler, NOrman S. "Changes in.Meaning During Psychotherapy as Measured by the Semantic Differential," Journal of Counseling Psychology, 8:105-111, Summer, 1961. {/Festinger, Leon and James M. Carlsmith. ”Cognitive Consequences of Forced Compliance,” Journal of Abnormal and Social Psychology, 58:203-210, 1959. Grigg, Austin E. ”A‘Validity Study of the Semantic Differential Technique," Journal of Clinical Psychology, 15:179-181, April, 1959. L///He1mreich, Robert and Barry E. Collins. ”Studies in Forced Compliance: Commitment and Magnitude of Inducement to Comply as Determinants of Opinion Change," Journal of Personality and Social Psycholpgy, 10:75-81, NOvember, 1968. Husek, T. R., and M. C. Wittrock. "The Dimensions of Attitudes Toward Teachers as Measured by the Semantic Differential," Journal of Educational Psychology, 53:209-213, October, 1962. V/Janis, Irving L. and J. Barnard Gilmore. "The Influence of Incentive Conditions on the Success of Role Playing in.Modifying Attitude," Journal of Personality and Social Psychology, 1:17—27, 1965 Kelly, Jane A. and Leon H. Levy. ”The Discriminability of Concepts Differentiated by Means of the Semantic Differencial," Educational and Psychological Measurement, 21:53-57, Spring, 1961. 71 L/Kiesler, Charles A., Michael S. Pallack, and David E. Kanouse. "Interactive Effects of Commitment and Dissonance," Journal of Personality and Social Psychology, 8:331-338, March, 1968. Sensening, John and Jack W; Brehm. ”Attitude Change From an Implied Threat to Attitudinal Freedom," Journal of Personality and Social Psychology, 8:324-330, April, 1968. y/Sereno, Kenneth K. ”Ego-Involvement, High Source Credibility, and Response to a Belief—Discrepant Communication,” Speech Monographs, 35:476-481, November, 1968. Sereno, Kenneth K. "Ego-Involvement: A Neglected Variable in Speech Communication Research," The anrterlnyournal of Speech, 55:69-77, February, 1969. Smith, Raymond G. ”Validation of 3 Semantic Differential," Speech Menographs, 30:50-55, March, 1963. Abstracts Rand, Melvin.A. ”An Empirical Comparison of Sherif's Social Judgment Approach and Festinger's Dissonance Theory at Their Points of Contrast Ego-Involvement and Discrepancy of Communication,” (unpublished Doctoral Dissertation, The University of Oklahoma, 1967), Dissertation Abstracts, 29A.l:1128-1129A, 1967. Sherman, Steven James. "The Effects of Importance, Choice, and Change Under Forced Compliance," Dissertation Abstracts, 28:2777A, 1968. Unpublished.Materials Coleman, Donavon E. "The Classical Conditioning of Attitudes Toward Selected Educational Concepts," unpublished Doctoral Dissertation, University of South Dakota, Vermillion, 1966. ///Hunt, Martin F. ”Open- and ClosedAMindedness and Self-Persuasion: Incentive y§_Dissonance Theory," unpublished Doctoral Dissertation, Michigan State University, East Lansing, 1969. Newspapers Capital Times GMadison, Wisconsin), September 23, 1970. Pe tomist (Wisconsin State University-Superior, Wisconsin), April 29, 970. November 19, 1970. APPENDICES APPENDIX A THE SEMANTIC DIFFERENTIAL EMPLOYED IN THIS INVESTIGATION 72 Your Name INSTRUCTIONS The purpose of this study is to measure the meanings of certain things to various people by having them judge them against a series of descrip- tive scales. In taking this test, please make your judgments on the basis of what these things mean to ygg, On each page of this booklet, you will find a different concept to be judged and beneath it a set of scales. You are to rate the concept on each of these scales in order. Here is how you are to use these scales: If you feel that the concept at the top of the page is vggy closely related to one end of the scale, you should place your check~mafk as follows: good X : : : : : : bad or good : : : : : : X bad If you feel that the concept is quite closely related to one or the other end of the scale (but not extremely), you should place your check-mark as follows: rugged : X : : : : : delicate or rugged : : : : : X : delicate If the concept seems only slightly related to one side as Opposed to the other side (but is not really neutral), you should check as follows: hot : : X : : : : cold or hot : : : : X : : cold The direction toward which you check, of course, depends upon which of the two ends of the scale seems most characteristic of the thing you're judging. If you consider the concept to be neutral on the scale, both sides of the scale equally associated with the concept, or if the scale is completely irrelevant, unrelated to the concept, then you should place your Check-mark in the middle space: bass : : : X : : : treble 73 INSTRUCTIONS_(Continued) IMPORTANT: (1) Please your check-marks in the middle of the space, not on the boundaries: : X (2) Be sure you check every scale for every concept-- DO NOT OMIT ANY. (3) Never put more than one check—mark on a Single scale. Sometimes you may feel as though you've had the same item before on the test. This will not be the case, so do not look back and forth through the items. Do not try to remember how you checked similar items earlier on the test. Make each item a separate and independent judgment. work at a fairly high speed through this test. Do not worry or puzzle over individual items. It is your first impressions, the immediate "feelings" about the items, that we want. On the other hand, please do not be careless, because we want your true impressions. NOTE: THERE.ARE NO "RIGHT" OR "WRONG" ANSWERS--WE WANT YOUR FIRST EMPRESSIONS. good small ugly hard S 0111‘ weak dirty low agitated tasty valuable green kind loud deep unpleasant 74 HIGH NON-RESIDENT TUITION bad large beautiful soft sweet strong clean high calm distasteful worthless red cruel soft shallow pleasant 75 HIGH NON-RESIDENT TUITION Page 2 black : : : : : : white bitter : : : : : : sweet sad : : : : : : happy dull : : : : : : sharp full : : : : : : empty peaceful : : : : : : ferocious heavy : : : : : : light sacred : : : : : : profane relaxed : : : : : : tense cowardly : : : : : : brave poor : : : : : : rich clear : : : : : : hazy hot : : : : : : cold thick : : : : : : thin awful : : : : : : nice dark : : : : : : bright bass : : : : : : treble HIGH NON-RESIDENT TUITION Page 3 rounded 76 fragrant honest active smooth fresh fast unfair rugged far sick narrow angular foul dishonest passive rough stale slow fair delicate near healthy wide 77 INCREASED FEES FOR NON-RESIDENT STUDENTS bad : : : : : : good large : : : : : : small beautiful : : : : : : ugly soft : : : : : : hard SOUI‘ I 2 2 2 2 2 SWEET. weak : : : : : : strong dirty : : : : : : clean low : : : F : : : high calm. : : : : : : agitated distasteful : : : : : : tasty valuable : : : : : : worthless green : : : : : : red kind : : : : : : cruel soft : : : : : : loud deep : : : : : : shallow unpleasant : : : : : : pleasant 78 INCREASED FEES FOR.NON-RESIDENT STUDENTS Page 2 black white bitter sweet sad haPPY sharp dull empty full peaceful ferocious heavy light profane sacred tense relaxed brave cowardly poor rich hazy clear cold hot thin thick nice awful bright dark treble bass 79 INCREASED FEES FOR NON-RESIDENT STUDENTS Page 3 angular rounded foul fragrant honest dishonest active passive rough smooth stale fresh fast slow unfair fair rugged delicate far near healthy sick narrow wide APPENDIX B DIRECTIONS GIVEN TO THE EXPERIMENTAL AND CONTROL GROUPS 80 PRE-TEST INFORMATION One of the graduate students is compiling statistical data on the basis of the semantic differential. He has asked me if you would be willing to fill out these forms to assist him in the statistical project. Your name is not needed, as this information will be used only for a statistical exercise. However, it would be appreciated if you would put your social security number in the Space provided for your name. These statistics will be analyzed on the computer, which has a program based on social security numbers. This should only take you a short length of time and will be of some assistance to the graduate student. 81 DIRECTIONS FOR SUBJECTS WITH LOW JUSTIFICATION AND HIGH COMMITMENT (IDENTIFYING TELEVISED USE OF STATEMENTS PUBLICLY) One of our graduate students is doing a study of the logic used by a given group. He has asked us to make a video-tape for use over university closed—circuit television and possibly over the local educa- tional television station. Other classes will be asked to record in a similar situation except over the radio. The graduate student has asked to have each of you prepare a three-minute speech supporting high non-resident tuition for state universities. It does not matter whether or not you believe in high non-resident tuition; he is interested only in illustrating the logic used by a given group. In other words, he wants you to formulate the best possible argument you can in support of a high non-resident tuition. I will give you 20 minutes to outline your speech and the argu- ments pertinent to your speech. Please do it on this paper and sign your name in the upper right—hand corner. Prior to starting your writing we would like to focus the camera on each of you, so that we may test the television equipment. Please look at the monitor. .At a later date you wdll be expected to record the speeches on video-tape at the television studio. For the purposes of the demonstration it will be necessary that you identify yourself and make your speech as forceful as possible. Those people who will see your speech will assume that you believe what you say. You do not have to participate if you choose not to do so; however, the student does need your help, and this project should not take over 35 minutes of our time today. Just prior to class time, one of the members associated with the experiment indicated that you may keep the outline of your speech until you are contacted to make the recording. Apparently they will be contacting you just as you leave class or in the very near future. I have been informed that the recording will be arranged at your con- venience. 82 DIRECTIONS FOR SUBJECTS WITH LOW JUSTIFICATION.AND LOW COMMITMENT (ANONYMOUS USE OF STATEMENTS PUBLICLY) One of our graduate students is doing a study of the logic used by a given group. He has asked us to participate in an anonymous recording for use over the university radio station and in individual classrooms. Other classes will be asked to record in a similar situation except over closed-circuit television. The graduate student has asked to have each of you prepare a three- minute speech supporting high non-resident tuition for state universities. It does not matter whether or not you believe in high non-resident tuition; he is interested only in illustrating the logic used by a given pro- fessional group. In other words, he wants you to formulate the best possible argument you can in.support of a high non-resident tuition. I will give you 20 minutes to outline your speech and the argu- ments pertinent to your speech. Please do it on this paper and Sign your name in the upper right-hand corner. The student making this study will randomly select most of you to go up to the radio station to record your speech. Please do not be concerned about signing your name. Once the random selections have been.made, you will no longer be identified as an individual. The name is merely to identify you so that you.may be requested to come and record your speech. Your speech will then be fUr- ther edited and spliced, so that any possibility of your being identified as an individual is non-existent. Ybu do not have to participate if you choose not to do so; however, the student does need your help, and this project Should not take over 35 minutes of our time today. Just prior to class time, one of the members associated with the experiment indicated that you may keep the outline of your speech until you are contacted to make the recording. Apparently they will be con- tacting you just as you leave the class or in the very near future. Also, I have been informed that the recording will be arranged at your con- venience. 83 DIRECTIONS FOR SUBJECTS WITH HIGH JUSTIFICATION AND HIGH COMMITMENT (IDENTIFYING TELEVISED USE OF STATEMENTS PUBLICLY) A.member of the staff and some of his students are involved in a research project. They have asked us to make a video-tape for use over university closed-circuit television and possibly over the local educa- tional television station. Other classes will be asked to record in a shmilar situation except over the radio. The staff member and students have chosen to have each of you prepare a three-minute speech supporting high non-resident tuition for state universities. It does not matter whether or not you believe in high non-resident tuition; they are interested only in selecting certain statements from your speeches to demonstrate the logic used by a given group. In other words, they want you to formulate the best possible argu- ment you can in support of high non-resident tuition. I will give you 20 minutes to outline your speech and the argu- ments pertinent to your speech. Please do it on this paper and Sign your name in the upper right-hand corner. Prior to starting your writing, we would like to focus the camera on each of you. Please look at the monitor. .At a later date you will be expected to record the Speeches on video-tape at the television studio. For the purposes of the demonstra- tion, it will be necessary that you identify yourself and make your speech as forceful as possible. Those peOple who will see your speech will assume that you believe what you say. Ybu do not have to participate if you choose not to do so; however, I would like to encourage you to consider the project. It is being done by one of our staff members in conjunction with some of his students for a rather well-known organization. They h0pe to produce both the televised and the audio recordings into a demonstration model to be used in classes throughout the country for problem solving and other logic oriented projects. .Also, it would appear that as a result of this project, they will be able to gather some new information about how one structures his mind in a cognitive sense. If the group doing the project is able to pro- duce the results they expect, your time will be well spent because your efforts will lead to new knowledge in cognitive structure, as well as to the develOpment of a major learning unit. 84 DIRECTIONS FOR SUBJECTS WITH HIGH JUSTIFICATION AND HIGH COMMITMENT (IDENTIFYING TELEVISED USE OF STATEMENTS PUBLICLY) Page 2 Just prior to class time, one of the members associated with the experiment indicated that you may keep the outline of your speech until you are contacted to make the recording. Apparently they will be con- tacting you just as you leave class or in the very near future. Also, I have been informed that the recording will be arranged at your con- venience. 85 DIRECTIONS FOR SUBJECTS WITH HIGH JUSTIFICATION AND LOW C(MVII'IMENT (ANONYMOUS USE OF STATEMENTS PUBLICLY) A member of the staff and some of his students have asked us to participate in an anonymous recording for use over the university radio station and in individual classrooms for the purpose of studying the logic used by a given group. Other classes will be asked to record in a similar situation except over closed-circuit television. The staff member and students have chosen to have each of you prepare a three-minute speech supporting high non-resident tuition for state universities. It does not matter whether or not you believe in high non-resident tuition; they are interested only in selecting certain statements from your speeches to demonstrate the logic used by a given group. In other words, they want you to formulate the best argument you can for high non-resident tuition. I will give you.20 minutes to outline the speech and the argu- ments pertinent to your speech. Please do it on this paper and Sign your name in the upper right-hand corner. Those making this study will randomly select most of you to go up to the radio station to record your Speech. Please do not be concerned about signing your name. The name is merely to identify you so that they may request that you come and record your Speech. YOur speech will then be further edited and spliced so that any possibility of your being identified with the speech is non- existent. YOu do not have to participate if you choose not to do so; however, I would like to encourage you to consider the project. It is being done by one of our staff members in conjunction with some of his students for a rather well-known organization. They hope to produce both the televised and the audio recordings into a demonstration model to be used in classes throughout the country for problem solving and other logic oriented projects. .Also, it would appear that as a result of this project, they will be able to gather some new information about how one structures his mind in a cognitive sense. If the group doing the project is able to produce the results they expect, your time will be well spent because your efforts will lead to new knowledge in cog- nitive structure, as well as to the development of a major learning unit. 86 DIRECTIONS FOR SUBJECTS WITH HIGH JUSTIFICATION AND LOW CCWIITMENT (ANONYMOUS USE OF STATEMENTS PUBLICLY) Page 2 Just prior to class time, one of the members associated with the experiment indicated that you.may keep the outline of your speech until you are contacted to make the recording. Apparently they will be con- tacting you just as you leave class or in the very near future. Also, I have been informed that the recording will be arranged at your con- venience. 87 DIRECTIONS FOR CONTROL GROUP An experiment is being conducted to study cognitive structure, as well as the logic used by a given group. The following have been randomly selected to make an anonymous radio recording: XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX XXXXXXXXXXXXXXX The remainder of you are asked to participate in an identified television recording. The people doing the survey would like to have each of you prepare a three-minute speech either supporting or Opposing high tuition and fees for non-resident students in Wisconsin institutions of higher education. I will give you 20 minutes to outline your Speech and the argu- ments pertinent to it. Please organize your speech on this paper and sign your name in the upper right-hand corner. Remember that you.may either Oppose or support high non-resident tuition and fees, i.e., argue from your own point of view. .At a later time YOU‘Will be requested to record your speech. YOu do not have to participate in the experiment if you choose not to do so; however, your effOrts will be greatly appreciated and ultimately very useful. Those of you who are to be televised should now stand up one at a time, so that we can determine the best camera angles to use when your Speech is actually recorded. Just prior to class time, one of the members associated with the experiment indicated that you may keep the outline of your Speech until you are contacted to make the recording. Apparently they will be con- tacting you just as you leave class or in the very near future. Also, I have been informed that the recording will be arranged at your con- venience. 88 POST-TEST INFORMATION I have another favor to ask of all of you. You have just taken time in preparation for engaging in a publicly televised Speech (anonymously recorded radio speech). Previously, I had asked you to help one of our graduate students in a statis- tical project. Unfortunately, after the data was placed on the computer, the tapes were erased and all data was lost. He has asked us to fill out the semantic differential again. As we have already lost some of our class time, we might as well assist the fellow in his project. HOwever, this time if he has problems with losing his data, we will ask him to collect it in another class. APPENDIX C A DESCRIPTION OF THE FORTRAN IV PROGRAM DEVELOPED BY JEREMY FINN AND WIFIED BY DAVID WRIGHT 89 Jeremy D. Finn's Multivariance - Univariate and Multivariate Analysis of Variance and Covariance: A FORTRAN IV Program Occasional Paper No. I7 Office of Research Consultation School for Advanced Studies College of Education Michigan State University March 1970 DUPLICATED WITH PERMISSION OF THE OFFICE OF RESEARCH CONSULTATION, SCHOOL FOR ADVANCED STUDIES, COLLEGE OF EDUCATION, MICHIGAN STATE UNIVERSITY. 90 Multivariance - Version 4 June 1968 by Jeremy D. Finn, Department of Educational Psychology, State University of New York at Buffalo Modified for the Michigan State University CDC 3600 and 6500 computer systems by David J. Wright, Office of Research Consultation. This description contains excerpts from the complete Multivariance docu- mentation by Jeremy Finn, available on a limited basis at the Office of Research Consultation, 201 Erickson Hall. For ease in cross reference, page numbers and references are the same in both documents. TABLE OF CONTENTS Pages Program description . . . . . . . . . . . . . . l- 8 Preparation of a problem run . . . . . . . . . . . . . . . . . 9-19 Systems control cards . . . . . . . . . . . . . . . . . . 9-10 Program control cards and data . . . . . . . . . . . . . 11-19 91 MULTIVARIANCE A Generalized Univariate and Multivariate Analysis of Variance, Covariance, and Regression Program1 This and the following chapters describe a computer program which is now in operation at the State University of New York at Buffalo, and elsewhere, which will perform much of the multivariate analysis reported in preceding sections. Multivariance will perform univariate and multivariate linear estimation and tests of hypotheses for any crossed and/or nested design, with or without concomitant variables. The number of observations in the subclasses may be equal, proportional, or disprOportionate. The latter includes the extreme case of unequal group sizes involving null subclasses, such as might arise in the application of incomplete experimental designs. The program performs an exact least-squares analysis by the method described by Bock (1963). It is logically divided into three phases: input, estimation, and what has been termed "analysis." The input phase allows for six possible forms of data, four con- trolled by variable format cards. The data may be punched as: 1This chapter is a modified version of a paper by the same name presented at the annual meeting of SHARE (XXX), Houston, Texas, February 29, 1968. The version of Multivariancc presented in this manual was written and tested with the cooperation of the Computing Center of the State University of New York at Buffalo. The Center is partially supported by NIH Grant FR-00126 and NSF Grant GP—7318. Assistance with the programming of this version was provided by Mr. Fred Hockersmith. Several of the matrix subroutines were adopted and/or revised from those produced by the IBM Corporation (1966), and from Bock and Peterson (1967). The chi-square and F probability routines were written by Mr. Richard J. Sherin and appear in Clyde, Cramer, and Sherin (1966). Computations for the age x sex repeated measures design were performed by Mr. Roger Koehler. 92 MULTIVARIANCE _ 2 _ 1. Raw unsorted data, each observation with its own cell identification information. 2. Raw data, sorted by cells, each cell with its own header card. 3. Raw data, no header cards, in order by cells. 4, 5. Raw data to be read from an independently prepared binary tape. 6. Means, variance-covariance matrix, and cell frequen- cies. The last Option allows for reanalysis of data which may have been ‘ presented only in summary form, such as might be found in a text or journal article. The subclass frequencies may be highly unequal, and within the limits of the computer are not restricted in magnitude. No account need be made of null groups; the program will automatically detect them. Options are available for reading the data from non- system binary or BCD input devices. A variety of common data transformations is provided. In addition, the program will accept a matrix transformation which can be used to obtain linear combinations of the original variates. The transformation matrix may itself be automatically generated, for the analysis of repeated measures designs. A program option allows for the orthonormalization of the transformation matrix. After transformations, the observed means and standard deviations for each subclass are displayed. The estimation and analysis phases are based entirely on the specification of single-degree-of-freedom planned contrasts. Rather 93 MULTIVARIANCE - 3 - than placing artificial restrictions on the sums of the group effects, Multivariance provides a solution for the model of deficient rank by having the user determine linear combinations of the group membership effects which are of interest to him. Five common sets of between-group contrasts are available, includ- ing orthogonal polynomials. Others include the comparison of all experimental groups to a control, comparing each subgroup to the mean of the subgroups, and Helmert contrasts. In addition, the program will accept arbitrary contrast matrices constructed by the user, for problems for which contrasts other than the five provided are of interest. The designation of the effects is through the use of coded "symbolic contrast vectors" which are described in the following sections. The estimation phase of the program will estimate the magni- tude of the effects and their standard errors. In addition, subclass means and residuals may be estimated, based upon the fitting of a model of user-determined rank. The observed and estimated subclass means may be combined across subclasses as desired. Means may be estimated for null subclasses when this is apprOpriate. The error sum of squares and cross products are estimated, and are adjusted to yield the variances and correlations of the variates. This estimate of the population sum-of-squares and cross-products may be either the within-group term, the residual sum-of-products after fitting a given model to the data, or special effects which are user-determined. This feature allows for the analysis of any fixed, random, or mixed effects design. Only one of the possible error terms may be employed 94 HULTIVARIANCE - 4 _ in any given run however, so that multiple runs are necessary for testing all effects in a model where more than one error term is needed. The analysis phase may be repeated any number of times. It allows the user to select subsets of variables and covariates from his original input set and to perform the appropriate analyses. Designating more than one variable as a criterion measure will cause the program to.consider them simultaneously and to apply appropriate multivariate test criteria. If some of the variables selected are designated as covariates, the program will perform a regression analysis prior to the analysis of covariance to determine the rela- tionship between them and the dependent measures. The regression coefficients in raw and standardized form and their standard errors are estimated. If between-group contrasts, cell means, and residuals have been estimated, they will be adjusted for covariates and re- estimated. The partial correlations among the dependent measures, the adjusted variances and standard deviations, are calculated and displayed. Tests of hypotheses concerning relationships between the two sets of variables are provided in three forms. The first of these is a univariate and step-down multiple correlation analysis, to determine the relationship between the independent variables and the individual dependent measures. Second, a stepwise univariate and multivariate multiple regression analysis is performed to deter- mine the effects of the individual independent variables (or sets of independent variables; i.e. predictors may be entered into the 95 ' MULTIVARIANCE - 5 - regression equations singly or in groups). In all cases, the order of both sets of variables is determined by the user prior to the computer run. A third regression option is the calculation of the canonical correlations, the corresponding raw and standardized weights, the percent of variation in the dependent measures accounted for by each of the correlations, and finally tests of significance of them. This is the extent of the regression portion of the Multi- variance program. The program may be used to perform specific individual analyses. These include the simpler univariate analysis, the estimation of between-group effects, multivariate regression analysis, or just canonical correlation analysis by itself. If subjects have been grouped in a sampling design, the program will proceed with the appropriate orthogonal or exact non- orthogonal analysis of variance (or covariance, if covariates had been indicated). The contrasts established earlier are grouped for tests of hypotheses according to the user's desires. For each contrast or each factor in the model, the program will perform uni- variate and/or multivariate tests of significance, and the step-down analysis. This latter feature is of particular value for the analysis of repeated measures designs. If the user desires, a discriminant analysis may be performed for each contrast or set of contrasts. The variance of the dis- criminant function and the percentage of between-group variation attributable to it are computed by the program. In addition, the raw and standardized discriminant function weights are calculated, 96 HULTIVARIANCE - 6 - and the three tests of significance due to Roy, Hotelling, and Bartlett, are carried out. Finally, if the program is being employed to analyze a non- orthogonal design, the user may wish to reorder the between-group effects for subsequent hypothesis testing. Again the order must be determined by the user on an a priori basis. The program will easily allow for this reordering. An additional feature is of value for the analysis of incomplete designs. It often happens that in a complex design, or even simple designs which are based on survey data, a number of subgroups have no Observations. It is therefore necessary when choosing a set of contrasts to be certain that they are estmnable. The inestimable terms are frequently interactions and often difficult to locate by inspection. The current version of the Multivariance program includes the identification and loca- tion of such terms so that they may be removed from the model by the user. An annotated list of the output available from Multivariance follows the user's manual. The main program and all of its subroutines are coded in Fortran IV. Double precision is employed wherever accuracy may become a concern. Versions of the program will readily work on most 32K-word computers and larger. Instructions for adapting the program to s new system follow this manual. The programming technique which was used is somewhat unique. All labels, data, data matrices, lintermediate, and final results are stored within a single singly- subscripted long vector within the program. Within this vector, all 97 MULTIVARIANCE _ 7 - data are packed. That is, there are no unused core locations between the rows or columns of the matrices of a particular problem-run. In addition, there are no unused locations between matrices. This necessitates that the address of every matrix and of its elements be variable and adjusted to the particular problem. Only the non- zero portions of triangular matrices and half of symmetric matrices are kept in storage. Both are stored in packed form by rows (i.e. a,l, a21, ‘22: a3}, a32, a33, aul, etc.). Only the principle elements of diagonal matrices are used, and are stored in vector form (i.e. all, ‘22: ..., an“). Matrix operations on large matrices, such as the orthonormalization of a potentially very large basis, are segmented so that only portions of the matrix reside in core at any one time. Elements corresponding to null subclasses are eliminated whenever possible. The result of this "controlled storage mode" is that Multi- variance has the flexibility to handle a variety of different sized problems. In each case, the amount of core needed is a function of the particular job. With all other parameters minimal (e.g. with 1 dependent variable), the program will accept problems of about 1000 subgroups of subjects, or, of about 100 dependent variables and co- variates combined in a problem with very few subgroups (e.g. l or 2 groups, or in a regression analysis), with a 32K machine. A typical large problem would be about 100 cells and 20 variables. Multivariance is programmed in sections (not links) so that portions of it may be easily removed from core, allowing more space for the data vector, and thus for larger prOblems. There are no checks built into the j98 MULTIVARIANCE - g - program for exceeding the available storage memory. Guidelines for determining capacity are provided however, following the user's manual. In addition a list of removable sections and the options they contain is provided. This version of Multivariance has incorporated within it a set of about 50 checks for errors that may be made in attempting to use the program. The user is referred to an apprOpriate point in this manual, which describes the source of the error. The program has been tested on a large number of problems. Yet, "bugs" are inevitable. The author would greatly appreciate being notified of any error that is discovered. 99 PREPARATION OF - PROBLEM RUN: The sequence of control cards and statistical data for ORDER 0? a problem is listed in this section. Detailed instruc- tions for preparing the control cards and statistical data are given in the following section. INPUT DECK a) Systems control cards 3) Program control cards and data, as described on pg“ Ill-l9. ' Interim 3600 Control Cards 3600 systems control cards as of May 1, 1970 are: 1. 3. S. PNC (Problem Number. Card) supplied by computer laboratory Job Card: 3.303, prob nc., FINN, -.ss, last name, first name. Equip Card: 330012. u - u. MT(1915) Load Card: 3 was, (.1 Run Card: . 3mm, n.ss, 3000 Program control cards and data as described on pages 11-19. 100 -10; 3mm - May 1, 1970 The present version of FINN's Multivariance Program is soseuhat sore lisited than the version described on the following pages. At this ties the following restrictions apply: In at c tion Card Col 28 0, blank, or 1 only, i.e., options 2 and 3 do not exist. Cola 35-36 Logical Unit Nuxber if data is on tape (binary 300). Unit 8 is advised. Cola 44, 68 These options do not exist at this time. Estiaasion Specificagigg Card Col 66 Options 1, 2, and 3 do not exist Beans he: ‘22. t a! Les; Card of a Job Coluens 75-80 FINISH rather than CONTINUE in 60-67 or SPIP in 60-63. 101 MULTIVARIANCE - 11 - PROGRAM CONTROL CARDS AND DATA: All necessary information is marked with *. All else is optional. All numbers are right adjusted. The page numbers accompanying the description of each parameter tell where more detailed information may be found concerning the parameter's use. PHASE I - INPUT Title Cards* Card 1* Cols. 1-60 Input Description Card* Cola. 1- 4* Number of variables in input set. (p. 22)‘3€?‘ 1 (34¢; Cole. 7— 8* Number of factors in the design. (p. 22) Akenzavrvv'<~f' Col. 12* Data form. (p. 22) l - raw, subclass identification on each observation * 2 - raw, with header cards, by cells 3 3 raw, without header cards, by cells 4 - mean, frequency. variance-covariance summary data. 5 - Independently prepared binary tape, without header cards. 6 - Independently prepared binary tape, with header cards. Col. 16* Number of variable format cards. Punch 0 for data forms 5 or 6. (p. 22) Col. 20 1 if transformations; blank otherwise. Optional. (p. 23) Cole. 21-24 Number of variables after transformations if different from columns 1-4. Optional. (p. 23) Col. 28 Blank or zero for no transformation matrix. 1 if trans. mat. is to be read. 2 if trans. mat. is to be generated with Symbolic Contrast Vectors. 3 to orthonormalizc a generated trans. mat. Optional. (p. 23) Cola. 29-32 Number of variables after transformation matrix if different from columns 21—24. Optional. (p. 23) 102 HULTIVARIANCE _ 12 - Cola. 35-36 1 if data is on user-supplied magnetic tape on unit 8. Blank for standard (card) input. Optional. (p. 23) Col. 40 l to punch summary data, blank otherwise. Information will be printed whether or not this Option is used. Optional. (p. 24) Col. 44 0 or blank to rewind optional input tape before run. 1 to avoid prior rewind. Optional. (p. 24) Col. 48 l to suppress computation of each subgroup's standard deviations. Blank otherwise. Optional. (p. 24) Factor Identification Card(s)* (p. 25) 6-character factor names followed by 4 digits giving number of levels for that factor. 8 factors per card. Up to 10 factors total. Comments Cards. Optional. (p. 26) Columns 1-80, comments as desired. Maximum number - 300. End-of—Comments Card* (p. 26) Cole. 1- 6 FINISH Variable Format Cardgs)* (p. 27) As required. Describe: Cell identification (Integer), then dependent variables and covariates (Floating Point) for data form 1. Dependent variables and covariates only for data forms 2 and 3. Cell identification (Integer) and one row of matrix of means (Floating Point) for data form 4. 103 MULTIVARIANCE Transformation Cards. Optional (p. 28) Gals. 1- 4 Location of resultant variable. Cole. 7- 8 Transformation code. Cola. 9—12 First variable to be transformed. Cola. 13-16 Second variable to be transformed. Cols. 17-20 Third variable to be transformed. Cols. 21~30 Constant to be used in transformation, ' including decimal point. 1 Blank Card at end, i; transformations are used. Variable Label Cards* (p. 32) 6-column alphameric labels, 13 to a card, as many labels as variables after all transformations. DAIA* Form 1 (p. 33) Each observation on one or more cards, variables preceded by cell identification. Last observation completely blank. Form 2 (p. 34) For each cell: a) Reader card: Cola. 1— 6 Number of observations in the cell (N033). Cols. 7-10 Level on first factor named on Factor Identification Card. Cola. 11-14 Level on second factor. Cols. 15-18 Level on third factor. etc. b) Observations for that cell (exactly N¢BS observations). Last data card: 1 blank card (following observations of last cell). 104 MULTIVARIANCE Form 3 (p. 35) a) Vector of cell frequencies. Format 13F6.0 b) All data cards, grouped by cells, in the same order as vector of frequencies. No headers. No end-of-data card. Form Q (p. 35) a) Matrix of cell means, 1 row at a time, each row preceded by cell identification. Last: 1 completely blank row. b) Vector of cell frequencies. Format 13F6.0. c) Format card for within—cell variance-covariance matrix. d) Pooled within-cell variance—covariance matrix, in packed form by rows. Form 5 (p. 37) Independently prepared binary tape. Total number of observations in columns 1—6 of a data card. Each observation preceded by a fixed point identification vector. Form 6 (p. 38) Independently prepared binary tape. Observations grouped by cells. Header cards in data deck. Last header card blank. Transformation Matrix. Optional. (p. 39) If column 28 of Input Description Card is.l a) Variable format card describing 1 row of the matrix. b) Matrix, one row at a time, as many rows as variables after the matrix is applied, each with as many elements as variables before the matrix is applied. PHASE II - ESTIMATION: 105 MULTIVARIANCE - 15 _ If column 28 of Input Description Card 15.; or_3 a) b) Design Card Cols 1-4 Number of factors in design on dependent vars. Cola 5-8 Number of levels of first factor Cola 9-12 Number of levels of second factor Cola 13-16 etc. (If any "number of levels" is negative, orthogonal polynomials may be used for that factor.) Col. 48 l to print original basis vectors as produced. Blank or zero otherwise. As many Symbolic Contrast Vectors as number of variables after transformation matrix. Estimation Specification Card* Cole. 1- 4* Rank of model for significance testing. Number of contrasts to be entered. (p. 41) Cole. 5~ 8 Rank of model for estimation--not to exceed the rank of model for significance testing. (p. 42) Col. 12* Error term for analysis of variance. (p. 43) Blank or zero = Within-cell. l - Residual. 2 - Special effects. Cola. 13-16 Degrees of freedom for error, if special effects are used - if column 12 - 2. Optional. (p. 44) Cole. l7~20 Number of alternative orders of effects to_ be established - other than the first. Optional. (p. 44) Cole. 23-24 Number of factors in the design for which arbitrary contrasts will be used Optional. (p. 44) Cole. 27-28 Number of factors in the design for which orthogonal polynomials will be used. Optional. (p. 451 Col. 32 l to have estimated cell means and residuals calculated and printed. Blank nrherwise. Optional. (p. 45) 106 ‘ MULTIVARIANCE - l6 - Col. 36 1 to have variance-covariance factors and correlations among the estimates calculated and printed. Blank other- wise. Optional. (p. 45) C01. 40 l to have basis vectors printed. Blank otherwise. Optional. (p. 45) C01. 44 Zero or blank for no combined means. 1 to have combined observed means and N's printed. 2 to have combined estimated means printed. 3 for both observed and estimated combined means. (For 1, 2, or 3 here, the following card must be a Means Key.) Optional. (p. 46) Means Keg. Optional. (p. 47) Numeric codes indicating which factors and inter- actions are to have means combined across levels and displayed. Codes separated by commas, ended with period. Comments following period. Free format. Any number of cards. Arbitrary Contrpgt Matrices. Optional. (p. 49) a) Factor name card, columns 1-6 as on Factor Identification Card. b) Variable format describing 1 row of contrast matrix. c) Contrast matrix, 1 row at a time. Each row contains n elements, where n is the number of subclasses for that factor. There are n-l such rows. Orthogonal Polynomials. Optional. (p. 50) Factor name card, columns 1-6 as on Factor Identification Card. PHASE III - ANALYSIS “107 MULTIVARIANCE _ 17 _ Symbolic Contrast Vectors* (p. 51) Optional if only regression analysis is desired. As many cards, or equivalent, as rank of the model for significance testing. Each SCV should indicate one or more complete contrasts involving all factors, even if only their grand means, and should have a contrast description. If there is no contrast label in columns 75-80, the first six non-blank characters following the contrast code will be used as the label. Basis Reordering Keys. Optional. (p. 59) As many keys as indicated in columns 17—20 of the Estimation Specification Card, each key beginning a new card and containing as many elements as the rank of tne model for significance testing, indicating a new order. Elements separated by commas, ended with a period. Comments following period. Free format. Any number of cards. The following set (A-D) may be repeated any number of times. A. Analysis Selection Card* Cole. 1- 4* Number of dependent variables. (NDEP) (p. 61) Cole. 5- 8 Number of covariates. (NCOV) Optional. (p. 61) Col. 12 1 if variables are to be deleted or re- arranged to obtain the desired sets. Covariates must be ordered last. Blank otherwise. Optional. (p. 61) 108 MULTIVARIANCE Cola. 13-16 Number of alternative basis orders to be run with this set of variables - other than the first. Optional. (p. 61) Col. 20 1 for principal components of correlation matrix. Blank otherwise. Optional. (p. 62) C01. 24 l to perform discriminant analysis. Blank otherwise. Optional. (p. 62) Col. 28 1 to print hypothesis mean products. Blank otherwise. Optional. (p. 62) Col. 32 1 to perform canonical correlation analysis, if there are covariates. Blank otherwise. Optional. (p. 62) C01. 36 1 if covariates are to be entered into regression by user's key. Blank to enter one-at-a—time. Ignore if no covariates. (p. 36) B. Variable Select Key. Optional. (p. 63) If column 12 on the Analysis Selection Card is 1, this key is entered with (NDEP + NCOV) elements, indicating the variable order. Elements separated by commas, ended with a period. Free format. Any number of cards. C. Covariate Crouping_§_1. Optional. (p. 64) If column 36 on the Analysis Selection Card is 1, . this key is entered, each element indicating the number of covariates to be added to the regression equations. Covariates taken successively in the order in which they have been selected by the Variable Select Key or by their order upon input. Elements separated by commas, ended in a period. Comments following period. Free format. Any number of cards. 109 HULTIVARIANCE D. H othesis Test Card 8 * (p. 65) One or more cards for the original basis order plus one or more for each alternative order indicated in columns 13-16 on the Analysis Selection Card. Indicates grouping of between~cell degrees of freedom for tests of hypotheses. Degree of freedom numbers separated by commas, ended with a period. Comments following period. Free format. Last Card of a Job.* (p. 67) Cols. 60-67 CONTINUE if data for another run follows immediately. Cols. 60-63 STOP if this is the end of the last run. In case of a job error, cards will be read until one of these codes is encountered. "ITI'LI’TIHMIIITIIIMII'IIIITS