MESSAGE MANIPULATIONS IN COMMUNICATION OF A COMPLEX POUTICAL ISSUE Thesis for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY BARBARA EVERITT BRYANT 1970 IIII!!!IfllyfllullflfljflwflflwfllIIIILIIIIIII if . Miehigan Sm: University This is to certify that the thesis entitled MESSAGE MANIPULATIONS IN COMMUNICATION OF A COMPLEX POLITICAL ISSUE presented bg Barbara Everitt Bryant has been accepted towards fulfillment of the requirements for Ph.D. . Communication ______ degree in — flamfl Major firofessor June 9, 1970 Date 0-169 (2. M, 118’ {Ir-”OW We: WES gov BM- ABSTRACT MESSAGE MANIPULATIONS IN COMMUNICATION OF A COMPLEX POLITICAL ISSUE By Barbara Everitt Bryant Persuasive messages about a complex political issue were constructed in various ways by manipulation of three message variables: Source Label (Republican/Neutral), Amount of Use of Slogan (No Slogan Repeat/Slogan Repeat), and Format (Question-Answer/Straight Descriptive). These produced eight (2x2x2) message versions. The purpose was to determine if changes in any of these variables would influence either attitude change toward the issue or comprehension of the message among voters who read messages containing the same content and arguments. Printed brochures with the eight message versions were utilized in a field survey of a statewide sample of registered voters in Ohio, 300 in the experimental group and SO in a control group. The issue of the messages was one which affected these voters in their real-life situations: state support of education by the 1969 Ohio legislature. Each voter in the sample was categorized on the basis Barbara Everitt Bryant of his recent past voting behavior as Republican, Democrat, or Ticket Splitter, rather than by the more customary self— designation categories of Republican, Democrat, and Inde— pendent. Ticket Splitters formed a broader group between strong Republicans and strong Democrats, only 36% of the Ticket Splitters self—identified as Independents. Subjects were interviewed in their homes to determine demographic characteristics, past voting behavior, and pre- reading attitude toward the issue. Attitude was measured with four statements, two worded positively and two worded negatively, to which subjects responded on a five—point agree— disagree scale. Subjects in the experimental group were given one of the eight message versions to read. A seven— question multiple choice test on facts of the issue was ad— ministered to experimental and control groups. Following this, those in the experimental group received a post-reading attitude test identical to the pre-reading test. The messages used produced favorable attitude change at a significant level, producing positive change in 51% of the sample. Attitudes for the eight groups which received different message versions ranged from means of 0.7 to 1.3 above pre—reading attitudes which had a mean of 13.1. The prediction was made and confirmed that valencing the source with a Republican label would result in more posi- tive attitude change among Republicans and less favorable attitude change among Democrats than if they received the same message from a neutral source. Democrats' attitudes Barbara Everitt Bryant dropped more than Republicans' attitudes gained as the result of the Republican label. The prediction that Ticket Splitters would be unaffected by source label, because it had no valence for them, was not confirmed. They moved more positively as the result of receiving a Republican rather than a neutrally- labeled message. Predictions that importance of the issue to the sub- jects, and repeated use of a slogan would enhance favorable attitude change were not confirmed. In fact, repetition of a slogan, though it produced no significant main effects, had a significant interaction effect with Source Label and Voting Behavior Type on attitude change and the result was less overall attitude change from repeated slogan versions. As predicted, the amount read of a message had a significant effect upon comprehension. Of interest was that those who read all of a message had more comprehension gain compared to those who read 2353 of it than those who read part of it had compared to those who read nothing. Impor- tance of the issue to the voter did not affect comprehension. The Question—Answer Format was hypothesized to in- crease learning of facts over the Straight Descriptive Format. However, Format showed no significant effect on comprehension. Correlates of attitude change toward the issue and comprehension of the message were summarized, as well as characteristics of the Ticket Splitter voters. Accepted by the faculty of the Department of Communication, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the Doctor of Philosophy degree. Director of Thesis Guidance Committee WM Chairman W flWJWM/ MUM- MESSAGE MANIPULATIONS IN COMMUNICATION OF A COMPLEX POLITICAL ISSUE by Barbara Everitt Bryant A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Communication 1970 Copyright by BARBARA EVERITT BRYANT 197I ACKNOWLEDGMENTS I have had personal contact with four generations of my family and all, directly or indirectly, have contrib— uted to the accomplishment this dissertation represents. Most immediately, my husband John H. Bryant is in large part responsible for giving me the motivation and the tangible and psychological support to pursue graduate study. My daughters, Linda and Lois, and my son, Randal, have dis— played interest and patience, knowing that some of the time for my study was at their expense. As they embark on their college years, I hope they will benefit from my experiences as I have received inspiration from the social awareness and lack of hypocrisy of their generation and the individual achievement and high good humor of the three of them person- ally. I thank my parents, Dorothy and William L. Everitt, for showing me that, since change is both necessary and in- evitable, education is a lifelong quest to be pursued both inside and outside the classroom. My parents, my parents- in-law Cordie and John P. Bryant, and before them Benjamin H. Everitt, have demonstrated so forcefully in their lives that age is an attitude not a number. I am the beneficiary of their youthful and active attitudes toward learning and living. ii In the research and writing of this dissertation, and more importantly in the research project which is its subject, I have had the supervision and personal tutorial help of my advisor, Dr. Verling C. Troldahl. I realize better in retrospect than I may have at the moment, how much he has done to prod me into thinking through a problem, recognizing what is important and what is extraneous. I thank him for his concern, his knowledge, his dynamic and imaginative approach to research. I also thank the members of my advisory committee, Dr. Bradley S. Greenberg, Dr. Randall P. Harrison, and Dr. John T. Gullahorn, from whose courses, informal conferences, and questioning I have gained much. Finally, this study was made possible by Market Opinion Research, Detroit, Michigan, which sponsored the field survey in Ohio utilizing professional interviewers to obtain the 350 interviews in this sample. I am indebted for the interviews and for the counsel and training given to me by Mr. Robert Teeter of Market Opinion Research and others of that organization's staff. iii TABLE OF CONTENTS -ACKNOWLEDGMENTS . . . . . . . . . . . . LIST OF TABLES. . . . . . . . . . . . . LIST OF FIGURES . . . . . . . . . . . . LIST OF APPENDICES. . . . . . . . . . . Chapter I THE RESEARCH PROBLEM . . . . . . Introduction: The Growing Use of Political Messages . . . . . . . . . . . The Need for Research. . . . . . Significance of the Study. . . . Background and Related Literature Study Design Considerations. . . Limitations. . . . . . . . . . . Hypotheses . . . . . . . . . . . II RESEARCH DESIGN AND METHODOLOGY. Purpose. . . . . . . . . . . . . Choice and Operationalization of General Study Design . . . . . . Sample . . . . . . . . . . . . . Data Collection. . . . . . . . . III FINDINGS . . . . . . . . . . . . Description of Sample. . . . . . Attitude Change Toward Issue . . Comprehension of Message Content Slogan Recall. . . . . . . . . . Further Explorations . . . . . . IV SUMMARY AND DISCUSSION . . . . . Summary. . . . . . . . . . . . . Discussion . . . . . . . . . . . Contributions to Theory. . . . . Implications for Future Research APPENDICES O O O O O O O O 0 O O O O O O BIBLIOGRAPHY O O O O O O O O O O O O O 0 iv Variables Page ii vii viii 52 52 53 66 67 69 7O 7O 7O 81 86 88 94 94 100 109 116 122 140 Table 10 ll 12 13 LIST OF TABLES Description of Sample . . . . . . . . . . . . . Incomplete and Non-returned Interviews Dropped from sample 0 O O O O O O O O O O O O O O O O 0 Mean Attitude Change for Message Manipulation verSionS. O O O O O O O O O O O O O O 0 O O O 0 Mean Attitude Change for Voting Behavior Type vs. Source Label of Message . . . . . . . . . . Analysis of Variance Summary Table for Depend- ent Variable, "Attitude Change Toward Issue" (Independent Variables: Voting Behavior Type, Source Label, Format, Amount of Slogan Use) . . Mean Attitude Change for Variables which Pro- duce Interaction Effect . . . . . . . . . . . . Analysis of Variance Summary Table for Depend- ent Variable, ”Attitude Change Toward Issue," Using No Slogan Repeat Messages . . . . . . . . Mean Comprehension of Message Content Score for Message Manipulation Versions . . . . . . . Mean Comprehension of Message Content by Amount of Readership and Importance of Issue to Voter. Analysis of Variance Summary Table for Depend- ent Variable, "Comprehension of Message Con- tent” (Independent Variables: Readership, Importance of Issue). . . . . . . . . . . . . . Analysis of Variance Summary Table for Depend- ent Variable, "Comprehension of Message Con— tent" (Independent Variables: Voting Behavior Type, Source Label, Format, Amount of Slogan Use). . . . . . . . . . . . . . . . . . . . . . Analysis of Variance Summary Table for Depend- ent Variable, ”Comprehension of Message Con— tent," with Education Level as Covariate. . . . Slogan Recall vs. Amount of Use of Slogan . . . Page 71 72 73 74 76 77 79 81 82 84 85 87 88 Table 14 15 16 Page Correlates of Attitude Change Toward Issue of Mssages. O O O 0 O O O O O O O O O O O O 0 O O 90 Correlates of Comprehension of Message Score. . 92 Mean Attitude Change for Voting Behavior Type vs. Source Label of Message (No Slogan Repeat verSionS) 0 0 O O O O 0 O O O O O O O O O O O O 101 vi Figure 10 ll 12 LIST OF FIGURES Heider p-o-x triads in balanced state. . . . . Newcomb A-B-X triangle in balanced state . . . Illustrative model of Cartwright and Harary graphs 0 O O O O O O O O O O O O O O O O O O O Feather Communication Model: Signed digraphs represent the eight possible communication structures in which all semicycles involving attitudinal and unit relations are completely balanced . . . . . . . . . . . . . . . . . . . Conditions set up by the experiment in the communication model. . . . . . . . . . . . . . Initial communication situation for receiver who receives communication identified with his own political party. . . . . . . . . . . . . . Balanced post-communication situation for re— ceiver who receives communication identified with his own political party . . . . . . . . . Initial communication situation for receiver who receives communication identified with the opposition political party . . . . . . . . Balanced post—communication situation for re— ceiver who receives communication identified with the opposition political party. . . . . . Post-communication situation for receiver who receives communication identified with his own political party. . . . . . . . . . . . . . Post-communication situation for Ticket Splitter O O O O 0 O O O O O O 0 O O O O O O 0 Post-communication situation for receiver who receives communication identified with oppo- sition political party . . . . . . . . . . . . vii Page 10 10 12 15 37 42 43 45 45 110 111 112 LIST OF APPENDICES Appendix I Message Version 1. . . . . . . . . II Message Version 2. . . . . . . . . III Message Version 3. . . . . . . . . IV Message Version 4. . . . . . . . . V Message Version 5. . . . . . . . . VI Message Version 6. . . . . . . . VII Message Version 7. . . . . . . . . VIII Message Version 8. . . . . . . . . IX Survey Questionnaire for Experimental Group. X Survey Questionnaire for Control Group . . . XI Questions from Preceding Survey Used in Dissertation Experiment. . . . . . . . . . . viii Page 122 123 124 125 126 127 128 129 130 135 138 CHAPTER ONE: THE RESEARCH PROBLEM I. INTRODUCTION: THE GROWING USE OF POLITICAL MESSAGES Political messages are intended to persuade whether they appear as advertisements, in speeches, in candidate interviews with reporters, in promotional literature and handouts, or in conversations among the supporters of an issue. As mass society becomes more and more complex, its issues become more complicated: problems of population con- trol, pollution, conservation, racism, educational oppor- tunity. An increasing number of the decisions which affect the quality of life in the whole society become political decisions, made by large numbers of voters either indirectly in choosing the candidate who becomes one of the decision makers of the society, or directly by deciding referendums on a broad variety of issue proposals. The voter is the target of political messages of fact and opinion directed at him by proponents and opponents of candidates, political parties and issues; by those who seek change and those who resist it. In the United States, political communication is a recurring phenomenon to which increasingly large amounts of money and media resources are allocated. "Spending in all campaigns for all offices at stake in 1968, from county commissioner to the presidency, totaled at least $300 mil- lion. That was a 50 percent increase over the $200 million spent in 1964, which was itself a record breaker."l A large percentage of this massive expenditure goes into the produc- tion and distribution of political messages. A generation ago political communications reached the interested and could be avoided by the apathetic.2’ 3 Today avoidance of political messages is hardly possible during major campaigns, though still can be managed at the level of local, county, and issue contests in which propo- nEnts of candidates or issues cannot afford extensive media use. Even here, increasing use of electronic media paralleled with growing and more varied use of print media are deliver- ing messages even to those who care little. Campaign empha— sis is shifting toward appealing to the voter directly through the mass media.4 lHerbert E. Alexander and Harold B. Meyers, "A Fi- nancial Landslide for the G.O.P.," Fortune, March, 1970, 104—105, 186-189. 2Paul F. Lazarsfeld, Bernard Berelson and Hazel Gaudet, The People's Choice——How the Voter Makes Up_His Mind in a Presidential Campaign (New York: Columbia Uni— versity Press, 19447. 3Bernard R. Berelson, Paul F. Lazarsfeld and Wil— liam N. McPhee, Voting: A Study of Opinion Formation in a Presidential Campaign (Chicago: University of Chicago Press, 1954). 4James M. Perry, The New Politics, The Expgnding Technology_of Political Manipulation (New York: Potter, 1968), p. 7. II. THE NEED FOR RESEARCH Little is known about how the voter reacts when he receives a political message about a candidate or issue. Does he learn the facts of the message, weigh those facts, identify the intent or bias of its source, allow the message to change or reinforce his attitudes about the issue or candi- date it concerns? Or does he interpret the message only in terms which do not upset the attitude he had, or voting decision he would have made, if the message had never reached him? This study explores some of these questions. Despite the large number of political messages is— sued in the United States each year, no field studies have been made in which political messages have been manipulated and tested for their differential effects on a sample drawn from a population of registered voters.5 There is, however, a significant body of research on attitude change, the re- sults primarily of experimental studies with student subjects. There is also a large body of survey and polling studies of the voting population in the U.S. These studies cover the period 1940 to date and identify variables in the audience for political messages, not message elements themselves. 5A literature search for articles relevant to polit— ical communication was made by the experimenter in the fol— lowing journals for the years l959-early 1970: Public Opinion Quarterly, American Political Science Review, Social Forces, Journal of Communication, Journalism Quarterly. The search also included books and articles referenced in any of the articles on political communication in the above. A key word search of theses from 1950 to 1970 by University Microfilms, Ann Arbor, Michigan, showed no references under the key words Message and Political and Communication; Message and Political; Manipulation and Political and Message, Com— munication. III. SIGNIFICANCE OF THE STUDY The significance of this study is that it attempts to relate existing knowledge of attitude change and the vot— ing audience to the important, but largely untested, field of political message effects. The study uses messages about a typical issue, manipulated and tested upon those whom the issue affects in real life. The issue itself, a bill provid- ing state support for education, is typical of many complex issues which are debated and acted upon by political bodies: it affects many; it contains a mixture of both widely and long agreed upon proposals plus new and highly controversial proposals. It is the end product of compromises; it is ex— pensive. §2fl2.0f its details are familiar to ppm; voters—— only specialists understand the whole package. Despite the multiple number of elements in the bill, it must ultimately be judged as a single unit by the voter. Thus the political messages used in this study are realistic compared to those typically administered in laboratory settings. The experimenter selected three message variables from among many possibilities; presented them to a statewide sample of 300 registered voters (plus 50 non—readers in a control group); and then measured comprehension and attitude change effects among readers. Additionally, voters were interviewed both before and after reading the messages to ascertain characteristics about them—~demographic, attitud— inal, and voting behavior——which might influence their per— ceptions of the messages. IV. BACKGROUND AND RELATED LITERATURE l. Theories of Attitude Change and Attitude Stability . Political communications have two goals: (1) to change the attitudes of those who do not agree with the com- municator's stand and (2) to reinforce the attitude of those who aiready agree with the communicator. Theories contributing to an understanding of atti- tude change and attitude stability are classified under the general heading of balance or consistency theories. All are in agreement that people change their attitudes to elim— inate some inconsistency and make no changes when cognitions are consistent.6 a. Congruity and Dissonance Consistency theories have been explicated in a num— ber of ways. Osgood and Tannenbaum7 say in their congruity principle that the existence of an incongruity directly gen— erates the pressure toward change. Their theory is worked out from the standpoint of a receiver who holds initial at— titudes toward both a source and a concept. Source and con— cept are then linked by an assertion (a message or statement). 6R. Brown, "The Principle of Consistency in Attitude Change," Social Psychology (New York: The Free Press, 1965), pp. 549—609. 7C. E. Osgood and P. H. Tannenbaum, "The Principle of Congruity in the Prediction of Attitude Change," Psycho- logical Review, 62 (1955), 42-55. If the assertion results in a situation which is incongruous to the receiver, his evaluation of both source and concept will change to make the situation congruous to him. The experimental study reported in this dissertation examines the initial attitude and attitude change of the receiver of a message toward the political issue of that message. Tannenbaum illustrates the application of the con— gruity theory to such political attitudes: The fundamental homeostatic mechanism attributed to congruity theory is perhaps most apparent in the generalization studies where changes in evaluation of an object of judgment occur without any direct manipulation of that object. It is more than likely that many of our "real-life" attitudes are formed and modified along just such lines. Political atti- tudes constitute a prime example. Most of us rarely have any direct contact or experience with a polit— ical personality, but we often develop quite intense attitudes for or against him as a result of the stands he assumes on a number of issues toward which we already have some well—defined attitudes. Once formed, such attitudes toward this politician become factors around which other opinions are developed as he continues to take positions pro or con a num- ber of novel political issues. To be sure, such opinion formation does not always take place in a complete vacuum of factual information, but quite .often such information considerations occupy a sec— ondary role. Turning from congruity to another approach, the con- sistency theory which has generated the most research is 8Percy Tannenbaum, "The Congruity Principle Revisit— ed: Studies in the Reduction, Induction and Generalization of Persuasion," Advances in Experimental Social Psychology, Vol. 3, ed. by Leonard Berkowitz (New York: Academic Press, 1967), p. 317. Festinger's Theory of Cognitive Dissonance,9 where dissonance is defined as nonfitting relations among cognitive elements. Festinger sees dissonance as the inevitable consequence of making a decision, its magnitude dependent upon the impor- tance of the decision and the comparative attractiveness and overlap of cognitive elements between the alternatives. Prior to making a decision an individual is in a state of conflict, :.a state in which Festinger and his colleagues10 have experi— mentally demonstrated he weighs alternatives fairly objec— -tively. After decision, and only if the decision commits --the individual to the chosen alternative, he is in an immedi— ate post-decision period of regret when dissonance is salient. :Within a short time, however, dissonance reduction processes :begin. It is these processes which bring about attitude change. More partiality and bias enter evaluations, bring— ing about a divergence in the attractiveness of alternatives with the chosen one becoming more favorable. There is a tendency to look more at consonant than dissonant informa- tion, though Festinger has found this is easily overcome if the dissonant information is potentially useful to the individual. What is the specific application of dissonance to 9Leon Festinger, A Theory of Cognitive Dissonance (Stanford: Stanford University Press, 19575. loLeon Festinger, Conflict, Decision, and Dissonance (Stanford: Stanford University Press, 1964). the subject of political attitude change? Festinger does not discuss how recently a decision must have been made, but requires only that there have been one. In regard to political attitudes, it appears that an individual who con- sistently and always votes with a particular political party is in a post-decision situation (even though he may have made the decision at the time of his first vote after age 21). The theory would then anticipate that he would attend more to consonant information from his own party, and avoid that from the opposition. The individual who votes a split Iticket, or who switches party vote from one election to an— other, is not in the post-decision state. Rather, as each election approaches, he is in the predecision or conflict situation where he weighs alternatives more impartially. He should thus be regarded by political proponents as a prime target for political messages. These two explications of balance or consistency theory have been described briefly because they bear on the shaping of political attitudes through political messages. A relevant communication model based on balance theory has been developed, however, which deals even more directly with the effects of communication about an issue upon the receivers of the communication. The model is that of N. T. Featherll’12 11N. T. Feather, "A Structural Balance Model of Com- munication Effects," Psychological Review, 71 (1964), 291—313. 12N. T. Feather, "A Structural Balance Approach to the Analysis of Communication Effects," Advances in Experi— mental Social Psychology, Vol. 3, ed. by Leonard Berkowitz (New York: Academic Press, 1967), pp. 100—165. and is built upon the early work of Heider13 and the elabora- tion of Heider's ideas through the use of graphs by Cartwright and Harary.l4 To understand the model it is first necessary to review the earlier work on which it is based. b. Heider The earliest explication of balance theory was made by Heider who distinguished two types of relations, attitude (evaluating or sentiment) and cognitive unit formation (sim— ilarity, possession, causality, proximity, belonging). The unit relations concept stems from Gestalt psychology. Atti— tude relations are liking, L (positive), and not liking ~L (negative). Unit relations are U (there is a relation) and ~U (there is no relation). Two entities, for example, two people, are in a bal— anced state if the relation between them is positive or nega— tive in all respects (all meanings of L or U). Heider or— ganized three entity relations, two people and an object or concept, into p-o-x triads from the point of view of person p. Here p and o are the persons and x is the object or con- cept. A balanced state exists if all three relations are positive or two are negative and one is positive. l3F. Heider, "Attitudes and Cognitive Organization," Journal of Psychology, 21 (1946), 107-112. l4Dorwin Cartwright and Frank Harary, "Structural Balance: A Generalization of Heider's Theory," Psycholog~ ical Review, Vol. 63, No. 5 (Sept., 1956), 277—293. 10 P —%> 0 P , O + + Fig. l.—-Heider p—o—x triads in balanced state If two relations of a potential triad exist, and there is pressure toward the establishment of a third relationship, Heider predicts the relationship "induced will be one which produces a balanced triad." Heider, like those who have followed him, says there is a tendency for cognitive units to achieve balanced states. If change is not possible, im— balance will produce tension. c. Newcomb Heider's p—o-x triangles are somewhat similar to the A—B—X triangles of Newcomb15 except that, whereas Heider's triads were from the point of view of person p, Newcomb con- siders the point of view of both persons, A and B. X A 'I‘ B x / A + Fig. 2.—-Newcomb A-B—X triangle in balanced state 15T. M. Newcomb, "An Approach to the Study of Com- municative Acts," Psychological Review, 60 (1953), 393—404. 11 Newcomb observes the interplay of forces among indi- viduals tending to consistencies among them for all of them. Newcomb assumes a "strain toward symmetry" with symmetry being the similarity of A's and 8’s orientation to X (an impersonal entity). Of interest in Newcomb's work is that he stresses the part played by communication in carrying out the dynamic operations of moving imbalanced situations toward balanced ones. d. Cartwright and Harary Cartwright and Harary extend Heider's theory through the use of graphs, directed graphs and signed graphs which permit them to deal with more than three elements. The communication model to be discussed shortly utilizes Cartwright and Harary's signed-directed graphs, called s—digraphs. In these, Heider's attitude relations are shown as arrows with both sign and direction. Here, if the receiver likes or has a favorable attitude toward the source, the graph shows a solid arrow in the appropriate direction. .4 S \ R If the receiver dislikes the source or considers the source counterattitudinal, a dashed arrow is used. 5(— —————————— R Heider's unit relations are shown as brackets with sign but not direction. Thus, if the communication is l2 perceived as belonging to or caused by the source, the re- lationship would be a positive unit relation shown by a solid bracket. c There can also be dissociative unit relations, as when the source denies responsibility for a communication, shown by dashed brackets. _q..&.- 5" ‘c Cartwright and Harary define paths, cycles, and semi- cycles to explain s—digraphs and the conditions for balance. Fig. 3 is an illustration which will be used to explain paths, cycles, semicycles and balanced structures. A <’ B > c ‘C _______ 4V0 Fig. 3.-—Illustrative model of Cartwright and Harary graphs The path from A to D could be AB), 3 or R), .C—II. A gyglg_is any path which may be taken through a set of rela— tions with beginning and end occurring at the same entity and with no relations traversed more than once. Thus a cycle can go from A and return to it through B and D or through C and D. A semicygle is a collection of lines 13 obtained by taking one from each pair in a cycle. Thus there are two semicycles in the ABD cycle: E), 8?, Wand AB, W, 9—2. A semicycle or cycle is balanced if the product of all of its lines is positive (lines have a value of +1 or -1). Thus here the ABD semicycles and cycle are balanced as all of the lines are positive. The ACD cycle, however, is imbalanced: (+1) x (+1) x (—l)=-1. Also, since the use of s-digraphs allows dealing with more than three elements, the semicycles and cycles ABCD are imbalanced: (+1) x (+1) x (+1) x (-l)=—1. An s-digraph is balanced if, and only if, all of its semicycles are positive. Cartwright and Harary also recognize the possibility that some unbalanced s-digraphs are more unbalanced than others. They define the degree of balance as the ratio of: Number of_positive semicycles Total number of semicycles in structure In the structure of Fig. 3 there are six semicycles: (1) 1353,33; (2%,???sz <3) T853701"; <4) xiiififmfi; (5) IA-‘B',—BU,%C,fi; and (6) AEJBE. Three of these are negative. Thus the degree of balance is 3/6=1/2. No effects are postulated according to the degree of balance but the con- cept is useful for comparing structures. Cartwright and Harary also allow for two other as— pects of partial balance in a structure. A structure is N-balanced if all cycles of length not exceeding N are 14 positive. In Fig. 3, for example, the structure is 3g; 3- balanced as one of the cycles of length three is not bal— anced. A structure is locally_balanced at any point p if all cycles passing through p are positive. In Fig. 3 the structure is pg; locally balanced at any of the four points as all have the unbalanced cycle ABCD passing through them. In more complex structures, however, it is possible to have some points of local balance in an overall unbalanced struc— ture. e. Feather's Communication Model Based Upon Balance Theory Feather uses Heider's attitude and unit relations, plus Cartwright and Harary's s-digraphs)as the basis for a communication model which is particularly appropriate for considering attitude change resulting from communication about an issue. This is exactly the situation for which the model was developed and for which it has been empiric— ally tested with laboratory groups. 1. Structure of the Model. The model is comprised of a source (S), a communication (C), and issue (I), and a receiver (R). The communication (C) takes some stand on the issue (I). Heider's relations, attitude and cognitive unit, link the four elements and can be diagrammed with s-digraphs, which represent the coqnitive structure of either source or receiver. Fig. 4 shows the eight possible communication structures involving the four elements of the model which ————— 15 would be completely balanced in both attitude and unit re- lations. In Fig. 4 the solid arrows are positive attitude relations, L, the dashed arrows are negative attitude rela- tions, ~L. The solid and dashed brackets are respectively positive unit relations, U, and negative unit relations, ~U. Fig. 4.--Feather Communication Model: Signed di- graphs represent the eight possible communication structures in which all semicycles involving attitudinal and unit re- lations are completely balanced. 6 It is not necessary that there be an attitude relation link— ing every pair of elements in the model in both directions. For example here, in Fig. 4, while the people 5 and R can hold attitudes toward the issue and toward each other, the issue cannot hold attitudes toward them. Thus no arrow points from I either to S or R. The communication, though inanimate, can express an attitude toward the issue and the l6Feather, "A Structural Balance Approach," p. 108. 16 C to I arrow is shown. The issue cannot reciprocate with an attitude about the communication. Similarly, there is not always a unit relation link— ing elements. In Fig. 4 the issue is not owned or possessed by S, R, or C nor are S, R, or C responsible for the exist- ence of the issue. Thus no brackets indicating a unit re- lation link them with I. Feather and his colleagues17 ran a series of experi- ments to demonstrate the model and found that subjects given information about some of the attitudes and asked to predict others tended to give responses which would result in a bal- anced model. They have made a series of findings related to use of the model, five of which will be listed here in brief though are described in full in the referenced article: (1) In the absence of contrary information, a re- ceiver tends to see a source as having the same attitude toward the issue as the communication he presents and as agreeing with the communication he presents. Even under conditions of source coercion, the receiver tends to see the communication as belonging to the source. This suggests that the receiver will see a positive unit relation linking source and communication. (2) When an interpersonal relationship is given, such as the attitude of source to receiver, it is a dominant l7Feather, "A Structural Balance Approach." l7 influence in predicting the opposite attitude relation of receiver to source, particularly when a balanced solution is impossible. (3) In an experiment, Feather and Jeffries found support for the hypotheses that when the stand of a commu- nication is in the same direction as the receiver's attitude toward the issue, the receiver will show both stronger agree- ment with the communication and more positive evaluation of the source as the stand comes closer to the receiver's attitude. Conversely, when the stand of a communication is opposite to the receiver’s attitude, the receiver will show both disagreement with the communication and a more negative evaluation of the source as the communication stand becomes more removed from the receiver's attitude. This is similar to the assimilation and contrast effect investi— gated by Hovland, Harvey, and Sherif.18 (4) Subjects rate the source of a more extreme com— munication as less valued, less credible, more potent and more active than the source of a moderate communication (extremity effect). Feather's experimental set up for this kind of finding is to use pre-communication questionnaires to determine subjects' attitudes toward an issue. He then 18C. I. Hovland, O. J. Harvey, and M. Sherif, "As— similation and Contrast Effects in Reactions to Communica— tion and Attitude Change," Basic Studies in Social Psychology, ed. by H. Proshansky and B. Seidenberg TNew York: Holt, Rinehart, and Winston, 1965), pp. 186-196. Fi———" 18 gives them communications from a source about the issue. Post-communication he measures the subjects' attitudes to- ward the source. Feather and Armstrong explored the communication model with strengths of attitudes (strong, weak, moderate), as well as their signs, specified and again found support for the Feather and Jeffries hypotheses and for the extremity effect. (5) Feather and Jeffries also found a bias toward evaluating the source positively. Thus they suggest other cognitive biases, such as extremity and positivity, besides the tendency toward balance, may impose limitations on bal— ance theory. Rosenberg and Abelson19 have identified, for example, that subjects seek the most pleasant solution to imbalance. 2. Limitations of the Model. Feather's communica- tion model based on balance theory does not account for strength of relations. Although he and Armstrong did ex- plore the model with one experiment involving attitude re- lations of varying strengths, Feather suggests that strength of relations is a subject for future research. The model at present deals only with positive and negative relations. 19M. J. Rosenberg and R. P. Abelson, "An Analysis of Cognitive Balancing," Attitude Change and Organization, ed. by C. I. Hovland and M. J. Rosenberg (New Haven: Yale University Press, 1960), pp. 112—163. 19 Of the theories, only congruity attempts to handle magnitude, and then only between receiver and source, and receiver and concept (the issue). Like the model, it has only a sign linking source and concept, and no evaluation by the re- ceiver of the assertion (comparable to the communication in the model). Feather's model has no way of handling the importance of the issue to source or receiver. Festinger has said the importance of cognitive elements involved in a decision is a factor influencing the magnitude of dissonance. Finally, the model does not suggest which particular mechanism an individual will choose to restore balance, or account for individual differences in tolerance of imbalance. In this it is like most explanations of balance theory. Although alternative modes of reducing inconsistency are suggested by all, little prediction is made of which mode will be used in a given situation, or by a particular type of individual. 3. Use of the Model. Even with these limitations, all of which Feather has suggested as fruitful areas for exploration, the model does offer the most suitable frame— work of any suggested to date for exploring the effects of a situation in which a receiver perceives a communication which emanates from a source and concerns an issue. Most 20Festinger, A Theory of Cognitive Dissonance. 20 of the other explanations of the drive toward balance or consistency do not lend to the handling of four elements, allowing for receiver attitudes toward the communication as well as toward the issue and the source. Furthermore, Feather21 has demonstrated that findings based on other theories (congruity, dissonance) can be interpreted in terms of his structural balance model. This model is also partic- ularly applicable to the forced exposure situation of this study. In a forced exposure situation some of the predic- tions of dissonance theory are not allowed to operate, for example, the prediction that an individual will actively avoid information which increases dissonance. 4. Dissertation Study Interests. This dissertation focuses interest on attitude change toward the issue, but also considers comprehension of the message and attitudes toward that communication. While a communication is intended to persuade, communicators would not anticipate that a single communication would change a voter's decision from one in opposition to a candidate or issue to one in favor. At best a single communication will only slightly increment or decrement an attitude. Hypotheses about attitude change will be based on Feather's model, but not tied exclusively to it. This is an exploratory study and interest also includes exploring 21Feather, "A Structural Balance Model." 21 strength of relations and importance of issues which the model is not equipped to handle. Before stating these hypotheses, however, considera— tion will be given to the receivers of a political message. Feather's model can be elaborated from the point of view of either the source or the receiver. Since interest here is in the effects of the message on the attitudes and com— prehension of the receiver, further use of the model will be from the viewpoint of the receiver. Much has been studied about receivers in the voting audience. Thus hypotheses about message effects must draw from this body of research as well as from balance theory. 2. Studies of the Voting Population in the U.S. Political messages are introduced into a population of voters. There has been considerable study of this popu— lation via both polls and surveys. From voter studies have come some predictors of voting behavior, but little study of message effects on this behavior. The only communication variable measured quantitatively has been "media usage.“ a. The Two Classic Study Groups The two classic study groups are Paul Lazarsfeld- Bernard Berelson and their colleagues, and the Survey Re— search Center of the University of Michigan. Lazarsfeld and Berelson examined the presidential campaigns in depth 22 in two communities, Erie County, Pa., in 194022 and Elmira, N. Y., in 1948.23 The Survey Research Center has made cam— paign studies of a nationwide sample of voters before and after national elections from 1948 to date.24’25’26’27’28 Both study groups have put considerable emphasis on demo— graphic characteristics of the voting population, but have also explored campaign interest, participation, media use, feelings of political efficacy, and candidate and issue per— ceptions. All reports on the voting population cite Voting and The American Voter, the principle publications of these two groups, because no studies as comprehensive have been 22Lazarsfeld, The People's Choice. 23 Berelson, Voting. 24Angus Campbell, Philip E. Converse, Warren E. Miller, and Donald E. Stokes, The American Voter (New York: John Wiley, 1960), also The American Voter (An Abridgment) (New York: John Wiley, 1964). Page numbers cited hereafter are from the Abridgment. 25Philip Converse, Angus Campbell, Warren E. Miller, and Donald Stokes, "Stability and Change in 1960: A Rein— stating Election," American Political Science Review, LV, No. 2 (June, 1961), 2694280. 26Philip E. Converse, Aage Clausen, and Warren E. Miller, "Electoral Myth and Reality: The 1964 Election," American Political Science Review, LIX, No. 2 (June, 1965), 321-336. 27Donald E. Stokes, "Some Dynamic Elements of Con— tests for the Presidency," American Political Science Review, LX, N00 1 (MarCh, 1966), 19—280 28Robert Bartley, "Did 1968 Win Forecast GOP Era?" Wall Street Journal, Sept. 10, 1969. 23 published since. It should be kept in mind that both books predate the use of TV to the extent this medium has been .Hused since 1964, particularly for political spot messages. Even the Survey Research Center updating articles do not touch upon the new omnipresence of media messages which changes the message environment of all potential voters and most particularly that of the low interest voter. ~What these classics demonstrate is that though the theory of democracy is based upon (1) a public having enough information to make a rational choice between policy alter- natives, and (2) an election presenting the electorate with recognizable party alternatives,29 in practice both ideas operate poorly. They operate poorly because in the public there is widespread lack of familiarity with issues and even among those who know something about issues there is only limited consensus as to which party advocates which policy.30 People tend to stay in politically homogeneous per- sonal environments where family, friends, fellow members of groups, and coworkers agree on perceptions of politics. Political discussion is most often with these people of like predisposition.31 This, of course, conforms to the idea of consistency theories which claims people seek balanced situations. 29Campbell, The American Voter, Chap. 17. 30Ibid., Chaps. 7 and 8. 31Berelson, Voting, Chap. 6. , I_ 24 Elections do vary in their results and so some atti— tude shift must take place in the total electorate from one election to another. The Survey Research Center group iden— tifies six dimensions of partisan attitude and demonstrates how these have influenced the outcome of elections.32’33 The dimensions are: (1) Attitude toward Democratic candi- date as a person; (2) Attitude toward Republican candidate as a person; (3) Attitudes toward the parties and candidates which relate to the benefit of various groups; (4) Attitudes toward the parties and candidates which relate to domestic policy; (5) Attitudes toward the parties and candidates which relate to foreign policy; and (6) Attitudes which relate to the general performance of the parties in national af- fairs. It is the information relating to these dimensions which can become the subjects of political messages. In line with the assumption of this study that the effect of a political message is at most to increment or decrement an attitude, Voting points out that "voting trends during the campaign are made up of a large number of small shifts over short distances of the political continuum." The Berelson group's study identified the homogene— ous political milieu in which most voters move, but also 32Campbell, The American Voter, Chap. 3. 33Stokes, "Dynamic Elements." 34Berelson, Votin , p. 33. 25 identified those voters most likely to change. They are the voters, of course, for whom their personal environments cease to be homogeneous, those who come under cross—pressures from conflicting demographic factor pulls, those who talk more with members of the opposition party. Their party choice will follow the weight assigned to issues (such as seeing domestic issues more important than foreign policy issues.) Appeals of the campaign, delivered by media exposure or personal contacts, become one cross-pressure on the voter. "But in general, media exposure gets out the vote and solid- ifies preferences. It crystallizes and reinforces more than it converts.“35 Both classic studies agree that those most exposed to the media are those who are more interested, more parti— san and least likely to change. Those most likely to change are those with lower interest and less issue knowledge. Neither the Berelson or the Survey Research Center groups see this as all bad: Low interest provides maneuvering room for polit— ical shifts necessary for a complex society in a period of rapid change . . . an important balance between action motivated by strong sentiments and action with little passion behind it is obtained by heterogeneity within the electorate.36 35 p. 248. 36 Berelson, Votin , Chaps. 2 and 7. Quotation from Ibid., pp. 314-315. 26 A characteristic of those low interest voters who determine political shifts is "lack of partisanship." Sur- vey Research Center finds the voter who identifies himself as an "independent" is less involved in politics, has a poorer knowledge of the issues of a campaign, a fainter image of its candidates, and relatively slight concern about its outcome as compared with those with some partisan attach- ment.37 This confirms what Berelson, Lazarsfeld and McPhee found in Elmira that both interest and voting intention re- late positively to degree of partisanship.38 b. Studies Using Public Opinion Poll Data Polls on candidate choice and issue opinions are the most popularly distributed of the voting population studies because their results are published in the mass media. The polls have become a valuable resource for docu— menting past voting behavior. Most of the national polling organizations in recent years have sent their old IBM cards to the Roper Public Opinion Research Center in Williamstown, Mass. 1. A Computer Simulation. In 1960 data at the Roper Public Opinion Research Center were used by the Simul— matics Corporation to give parameters for a computer 37Campbell, The American Voter, p. 83. 38Berelson, Votin , Chap. 2. 27 simulation which could make predictions of electoral behavior in the U.S.39 The simulation was based on the idea that candidates have alternatives upon which issues they stand. While the classic studies have shown other things such as social milieu and party of the voter are more important than issues in determining the voter's decision, these are things about which the candidate can do little. "He controls the issues he talks about. He has much less control of who he I“ is and who the voters are."4U For the simulation, data about how voters responded to past issues were treated as 52 issue clusters. The com- puter simulation was based on categorizing members of the electorate into 480 groups by socioeconomic status, size of community, regions of the country, religion, sex, eth- nicity, and whether they named party preference as Republi- can, Democrat or Independent. Data were used from 65 fairly comparable national surveys, 130,000 individuals. The num- ber of voter types in each state was estimated, to set up 48 synthetic states without local politics. Formulas stat- ing how voters would behave in a given situation were set up subjectively based on hypotheses about behavior under cross_pressure, and the data on what had happened previously 39Ithiel de Sola Pool, Robert Abelson and Samuel Popkin, Candidates, Issues and Strategies: A Computer Simu— lation of the 1960 and 1964 Presidential Elections—(Cambridge, Mass.: The M.I.T. Press, 1964). 40Ibid., pp. 8-9. 28 in similar situations. Party loyalty and habit measured by past Congressional vote was taken as the most important factor in determining the vote and other factors treated as leading merely to deviations from that base.41 In 1960 the only issue which had a significant nationwide net effect on the vote was religion. The simulation showed that one voter in ten shifted for a net cost to Kennedy of 2.3% of the total vote. However, while he lost in popular votes he gained in electoral votes on the religious issue. The product moment correlation between the Kennedy index on the simulation and the actual Kennedy vote in the election was .82. The simulation was based on data, none of which was newer than 1958. The correlation between the state-by—state result of 1958 polls and the actual outcome was .53. "The simulation, in short, portrayed trends that actually took place between the time the data were collected and election day two years later."42 The simulation model was tested again in the 1964 campaign and proved capable of simulating the outcome on the basis of cross-pressure theory as well as it had in 1960. 2. Studies of Party Choice. Another study which used poll data was that of V. 0. Key, Jr.43 who disagrees 4lIbid., p. 57. 421bid., p. 165. 43V. 0. Key, Jr., The Responsible Electorate-wRation— ality in Voting_l936—l960 (Cambridge, Mass.: Belknap Press of Harvard, 1966Y. 29 with the classic studies which depict voters as largely poorly informed on issues. Key did a secondary analysis of data accumulated from 1940-1960 by the Gallup Polls, Roper Polls and National Opin- ion Research Center Polls. He concluded that man is rational and the political institutions he has developed for election of the president are rational. He said that the electorate judges retrospectively, either approving or throwing out incumbent administrations. He demonstrated with more than 50 tables that those who switch parties on presidential vote from election to election, and those who remain with the same party choice, do so as the result of opinions on party handling of policies. Key categorized each voter as a Standpatter, a Switcher, or a New Voter according to whether his vote was for the same party or not as the last time he voted. Unlike the Independ— ent whom Survey Research Center finds poorly informed and disinterested, Key’s Switcher is no less interested, in- formed, or involved than the Standpatter. "Those who switch do so to support governmental policies or outlooks with which they agree, not because of subtle psychological or sociolog— ical peculiarities."44 The Switchers move toward the party whose Standpatters they resemble in their policy views. 44Ibid., p. ix. 4SIbid., p. 55. 30 The Standpatters do not have to behave as mugwumps to keep their consciences clear; they are already where they ought to be in the light of their policy attitudes.46 Unfortunately, no data about the Switchers have been studied below the presidential vote level and Key, who an- alyzed them, died in 1963. Goldberg47 agrees with Key that the voter is not merely the victim of sociological and psychological factors. He explains why party choice remaining the same as for the parents for three—fourths of all voters is a rational choice for them, and why switching allegiance is rational for the rest. The offspring will retain the party identification of the parent if the norms proffered by the parent are sub- stantively rational for the self-interest of the offspring. c. Professional Political Management Studies In recent years a new type of polling, in—depth image and issue polling, has come into use in political campaigning. Its use is part of what Perry48 describes as "the new politics": There are two essential ingredients of the new poli- tics. One is that appeals should be made directly to the voters through the mass media. The other is 46Ibid., p. 53. 47Arthur S. Goldberg, "Social Determinism and Ration~ ality as Bases of Party Identification," American Political Science Review, LXIII, No. 1 (March, 1969), 5—25. 48Perry, The New Politics. 31 that the techniques used to make these appeals-- polling, computers, television, direct mail--should be sophisticated and scientific.49 In-depth polls tell the candidate what are rele- vant issue structures, his standing with the public (the distance to the goal); and polls suggest the campaign pledges and requirements needed to develop candidate image. The new politics is practiced by the professional political management organizations which link social science with campaign techniques. One such organization is Market Opinion Research of Detroit-—predominantly a polling organ— ization—-which has handled polling for gubernatorial, sena- torial and congressional candidates in Michigan, Ohio, Indi— ana and elsewhere. Market Opinion Research credits the Ticket Splitter with being the decisive voter in contemporary campaigns. Whereas Survey Research Center, and most of the polls, iden— tify voters as party members or Independents, Market Opinion identifies them as "Strong Party Voters" or "Ticket Splitters.” Ticket Splitter is a behavioral definition based on past voting behavior rather than on self identification as having, or not having, partisan leanings. It refers to one who--on the same ballot--votes for candidates of more than one party. Market Opinion Research surveys show the Ticket 49Ibid., p. 7. SOIbid., p. 79. 32 Splitter may identify himself as a member of a party (T-S respondents in Ohio self—identified as 21.2% R, 38.9% D and 36.3% Independent in this study). He may be active in his party but wants to maintain his integrity as an independent voter and doesn't vote a straight party ticket. Demograph- ically, he is slightly better educated and has slightly higher income than the average of the electorate. He is younger but apt to be in his 30's rather than 20's. He perceives himself as extremely issue—oriented but in fact is influenced greatly by a candidate's personal characteristics. His media usage is high compared to the average of the electorate. Increasingly, the Ticket Splitters are the voters who deter- 51952 These statements are mine the outcome of elections. confirmed by the demographics in this sample. Ticket Splitters, like Key's Switchers, are the ones who cause the Republican candidate to win one election and the Democrat the subsequent one. They are decidedly the ones who result in a state's going Democratic for President and Republican for governor in the same election. As a practical campaign tactic, Market Opinion Re- search considers anyone who ticket splits, whatever the di— rection of his predominant partisan leaning, as more open 51Robert Teeter, Market Opinion Research, Detroit, Mich. Personal interview, April 2, 1969. 52Stephen Stockmeyer, Market Opinion Research, Detroit, Mich. Personal interview, April 2, 1969. 33 to persuasion than the Strong Party Voters. To have positive effects, this persuasion must take the form of relating the political candidate to real issues relevant to the voter's issue structure. d. Who Attends to Political Messages and Who Is Influenced by Them? Who attends to political messages? All of the studies agree that the consistently and highly partisan do--the strong Republican or Democrat of Survey Research Center, the Stand— patter of Key, the Strong Party Voter of Market Opinion Re— search-~the disinterested do not. Survey Research Center finds most of those who self classify as Independents are less interested than the partisans. Their apathy makes them a poor audience for political messages. While the highly and consistently partisan attend to political messages, they remain relatively unmoved by those not in accord with their own partisan direction. Thus it is the "others" in the electorate who are the audience possibly influenced. Why is there discrepancy between studies in the find— ings on political media usage, political activity and inter— est among those "others" who are not highly partisan? The answer is in definition. As Key has pointed out, and study questionnaires confirm,S3 Survey Research Center's Independent 53Survey Research Center, Pre—Election Study and Post-Election Study questionnaires 1964 Project 473; 1968 Project 45523. 34 is a product of its definition: "Independent" came to be defined as one of several categories of voters arrayed along a scale of party identification. Some persons proclaim themselves to be strong Democrats or strong Republicans. Others apprise themselves as not such strong partisans and thereby permit themselves to be categorized as weak Democrats or weak Republicans. Of those who claim that they are "Independents," some will concede that they lean Democratic or Republican. A few remain-- 5 to 10% of the electorate—-who stubbornly insist that they are "Independents" with no leanings in either partisan direction. This group of genuine "Independents" is not an impressive lot.54 Many Switchers and Ticket Splitters are thus cate- gorized as party members under a self—identification-by-party scheme. Market Research and Key define fewer people in the partisan categories. Their Ticket Splitters or Switchers form a broader band between the Republicans and Democrats. In the study which is the subject of this disserta- tion, only 36% of Ticket Splitters self—classified themselves as Independents. These may include the apathetic. The re- maining 64% of Ticket Splitters are the target audience pgobably most open to considering political communications. Before many elections they could be considered to be in the predecision stage where Festinger finds they will weigh alternatives objectively. Since this study is concerned with political commu- nication and particularly with those receivers whose attitudes 54Key, Responsible Electorate, p. 92. 35 may_be most affected by political communication, subjects in the survey sample will be classified as Republicans (those who vote straight or mostly Republican for state and national offices), Ticket Splitters (those who divide their vote be- tween parties) and Democrats (those who vote straight or mostly Democrat). Key's Switcher classification is not used because it has not been studied since 1963 and never below the presidential vote level. Also, evidence suggests many of the Switchers may be Ticket Splitters. One group not included in this study is the non-voter. As dropouts from the political process non-voters do not affect political decision making. In this sample of regis- tered voters they numbered 14 out of 321 usable interviews. Balance theory seems to provide a reasonable explana- tion for the findings on the behavior of the subsets of voters in the voting population. The Standpatter, Strong Party Voter is in a condition of consistency under no pressure for attitude change. The voters who experience imbalance-— perhaps because of disagreement with an incumbent adminis- tration, their own changing personal environment, or the salience of a particular issue--become the cross-pressured Switchers and Ticket Splitters who may be influenced by political messages. Two other studies further demonstrate the applicability of the balance theory in explaining voter behavior. 36 SigelSS demonstrated that partisans see their pre— ferred candidate so as not to experience imbalance. The images Republicans, Democrats and Independents had of an ideal president in 1960 Detroit sample correlated highly. The image Democrats had of ideal president correlated posi— tively with their image of Kennedy (f9: .697, p (.04); similarly for Republicans' ideal and Nixon image (/0 = .870, p (.02). But the Democrats' ideal and Nixon dropped to a correlation of fig: .273 and the Republicans with Kennedy to ,fl7z .164. Independents had less highly differentiated candidate images. Greenberg56 reported upon a survey in a local elec- tion situation in which voters who thought their own side would win were exposed to significantly more campaign infor- mation than those who thought their side would not win. More voters with consistent COgnitions used bulletins and leaflets easily screened by selective processes. V. STUDY DESIGN CONSIDERATIONS The absence of any published work on the effective— ness of political messages in achieving attitude change among SSRoberta S. Sigel, "Effect of Partisanship on the Perception of Political Candidates," Public Opinion Quarterly, XXVIII, No. 3 (Fall, 1964), 483-496. 56Bradley S. Greenberg, "Voting Intentions, Election Expectations and Exposure to Campaign Information," Journal of Communication, XV, No. 3 (Sept., 1965), 149—160. 37 registered voters prompted the design of this exploratory study of message effects. 1. Relation of Experimental Situation to Communication Model Feather's model is used to describe the cognitive structure of the receiver when he receives a message which makes a positive presentation of action on a political issue. A political body, described in the communication as the source of that action, is the institutional source of the message. Interest is in any change in attitude toward the issue and in comprehension of the facts presented about the issue. Fig. 5 illustrates the conditions set up by the ex- periment before the receiver attends to the message. In line with Feather's finding that the receiver perceives the communication as belonging to the source, a positive unit relation is shown between S and C. S + I A +- + C Fig. 5.—-Conditions set up by the experiment in the communication model: Positive unit relation between source and communication; positive stand of communication on issue; positive attitude of source toward issue. The solid arrow SIIrepresents the positive attitude 38 of the source toward the action it took on the issue. The solid arrow firepresents the positive presentation the com— munication makes of the issue. 2. Voting Types and Attitude Relations The findings of the voting population studies suggest a way to consider the attitude relation RS: prior to receipt of the communication. Voters are classified by past behavior as Republican, Democrat, or Ticket Splitter. When the source is identified by a political party label, positive or nega- tive attitudes are assumed for the Strong Party Voters and no R to S attitude for the Ticket Splitters. When the source is not identified with a political party, no R to S attitude is assumed for all voters. 3. Importance of Issue The choice of issue gave a way to determine the rela- tive importance of the issue to the voter-receiver. The issue, state support for education in Ohio for 1969—71, was one on which action had already been taken. The issue has a tax cost to all voters in the state but brings direct bene- fits only to those with a family member in school (from kin- dergarten to university). A pretest of questionnaires with a small sample in Toledo suggested that those with children in school attached more interest and importance to education and its financing. The elderly and those with no children tended to be disinterested rather than negative. Of course, it can be anticipated there are some negative attitudes toward 39 the issue in both groups simply because of tax cost. How— ever, because of the benefits to schools and students, it seemed a safe assumption that in the aggregate those with students in the immediate family could be classified as at— taching relatively_more importance to the issue as compared with the voters with no students. The information in the message would also have more utility for those with family members in school. This definition of more importance could also be carried to sub-issues of the issue, such as assuming that Roman Catholics and those with children in nonpublic schools would attach greater importance to state aid for nonpublic schools, those with university students would at— tach relatively more importance to financing higher educa— tion, etc. 4. Comprehension of Message One other fact is relevant to communication effects in a field situation: comprehension of the message. When experiments are run on college students, as most message effect studies have been to date, the experimenter does not need to concern himself about a wide comprehension—reading level differential. Such is not the case with a statewide sample of voters ranging from those with little education to those with graduate degrees. In fact, two subjects who were blind had to be dropped from the sample. Comprehension of message content was measured as a dependent variable in this dissertation study. 4O 5. Selection of Message Variables Finally, it was necessary to select the message vari- ables and determine which of these could be held constant and which manipulated. A survey of political messages showed variables frequently used were: choice of issue, use of party label, candidate vs. issue orientation, explicit de— tail vs. highly ambiguous presentation, emphasis on youth vs. maturity of a candidate, attack on the handling of an issue vs. offering specific proposals for solution of issue; redundancy, and an infinite number of style and format vari— ables. Determination was made to use a positive presenta— tion of action completed on a complex issue as the communi- cation content, constant across all message versions and to hold wording (hence facts and attitude presentation) con- stant. Eight message versions were produced by manipulating three message variables selected from among the possibili- ties in the categories above. Two of these variables were hypothesized to affect the attitude changes the message would produce and one was hypothesized to affect comprehension of the message. The variables were: (1) giving the source either a Political Party Label or a Neutral Label, (2) varym ing the amount of repetition of an ambiguous political slo— gan, and (3) presenting information in a question—answer vs. straight descriptive format. 41 VI. LIMITATIONS The limitations of the study were that it could only test a selected few message variables from among many avail- able. Three seemed to be the maximum given the size of sam- ple (300 in experimental group). The sample was drawn from Ohio voters and so findings are limited to statements about registered voters in Ohio in early 1970. The experimental situation involved forced exposure and no conclusions can be drawn about how the same messages would have been per— ceived under conditions of voluntary exposure. Before- reading and after-reading attitude measures were not far apart in time. Some respondents may have recalled their before-reading responses with the recall tending to bias their after-reading responses in the direction of not change ipg attitude. The message was written at a readability level requiring approximately completion of 8th grade. Thus some respondents, though not many, were excluded by their reading level from adequately attending to the message. Finally, the questionnaire used for data gathering came at the end of an interview which took from 50 to 80 minutes per respond- ent. There may thus have been some respondent fatigue. VII. HYPOTHESES l. Hypotheses About Attitude Change Hl A political party label attached to the source of a persuasive message will lead to more positive attitude change among voters favorable toward this party label, and less positive change among voters unfavorable to 42 this label, as compared to a neutral label message. Attitude change among voters for whom the party label has no valence will be unaffected by whether the source has a political party label or not. Interest here is in the attitude change difference according to whether the receiver reads a Republican or neu- tral label message containing the same content and arguments. The attitude resulting from the message with neutral source label is the base line for comparison, rather than initial attitude toward the issue. Fig. 5 has shown the conditions set up by the message experiment with a positive unit rela- tion between source and communication, and positive attitude relations between source and issue, and communication and issue. The voter who consistently votes Republican, upon receipt of a Republican message, will add a positive attitude toward the source to the structure set up in Fig. 5. Fig. 6 shows this addition. 4%»4 C Fig. 6.—-Initial communication situation for receiver who receives communication identified with his own political Party Clearly, the model for him is most easily balanced if he finds both the communication and the issue positive, Fig. 7. S .2 R N + + I + + C Fig. 7.--Balanced post-communication situation for receiver who receives communication identified with his own political party Making attitudes toward communication and issue positive, causes no pressure to change his attitude toward the source, though the positive nature of the communication might strength— en the already positive R attitude. As pointed out previous— ly, the model does not allow for strength of relations, and the +1 attitude, for example, cannot be graphed by the model's rules to a +2. It can only demonstrate the conditions for balance. From consideration of the balanced situation, the hypothesis is made that the positive attitude the receiver holds for the source, plus the positive attitude expressed by the communication about the issue results in a positive attitude toward the issue. One also must consider, however, that the receiver in this and all later situations may have come to the com- munication situation with 39mg previous attitude about the iSsue as well as the source. If this attitude were highly negative, the distance between receiver attitude and that 44 expressed in the communication would be so great that the contrast effect described by Hovland et al.57 or the extrem— ity effect Feather and Jeffries58 have demonstrated would occur. In the opposite direction, if the receiver came to the situation with a highly positive attitude toward the issue, there is less room for his attitude to become more positive. Thus it is possible that a previous attitude other than the one toward the source may modify the attitude change the model predicts on the basis of attending to the message. The form of presentation of the issue has been set up to try to make the attitude toward the issue more posi- tive whether the receiver reads either the neutral or Repub— lican label versions. However, it is important to state again that the best expectation for a single communication would be that it would increment an attitude. It is not likely to cause a major attitudinal shift. Turning now to the Democrat, for him the initial attitude toward a Republican—identified source is negative, Fig. 8 and he can readily achieve balance by regarding both issue and communication negatively, Fig. 9. Thus the per- ception of the source as counterattitudinal should counter— act or modify the positive presentation of the message. If on attending to the message, the receiver finds facts or ideas in it with which he agrees, his negative attitude 57Hovland, "Assimilation and Contrast." 58Feather, "A Structural Balance Approach," Expt. VI. 45 toward the communication should be reduced. If, for example, his attitude toward the communication shifted to positive the model would then be unbalanced to a degree of 4/7. The numbers in this degree of balance are obtained by consider- ing that the model is made up of seven semicycles. These follow paths connecting: (l) SRCS, (2) SICS, (3) RICR, (4) SRIS, (5) SRICS, (6) RSICR and (7) SIRCS. If the receiver regards the source and issue negatively, but becomes posi- tive toward the communication, semicycles (l), (3), (6) and (7) then have uneven numbers of negative relations. Sé__:___.R + C Fig. 8.-—Initial communication situation for receiver who receives communication identified with the opposition political party Fig. 9.—-Balanced post—communication situation for receiver who receives communication identified with the Opposition political party 46 It should be noted that in use of the model here semicycles and cycles are synonymous. Opposite attitudinal relations, such as source toward receiver, are not operationalized here, nor are combinations of unit and attitude relations linking elements. Only regarding the communication and issue nega- tively, however, will result in a balanced model, Fig. 9. Turn now to the Ticket Splitter who receives a Re- publican communication. No pre—communication attitude is predicted for him toward the source. Thus the model is bal- anced on the basis of the conditions set up by the experiment, with the receiver left to form his attitude toward the issue and the source on the basis of the attitude he first forms toward the communication itself. If he has a negative reac- tion, he balances the model by reacting as in Fig. 9. If he has a positive reaction, he balances the model by react— ing positively to source and issue as in Fig. 7. Feather and Jeffries59 have demonstrated a bias for evaluating a source positively. The message used in this experiment purposefully attempts to create positive attitudes toward the issue. Thus, it is anticipated that the tendency to balance the model positively as in Fig. 7 will outweigh the tendency to evaluate issue, communication and source nega- tively as in Fig. 9. This reaction will be the same for either the Republican or neutral label message. 59Feather, “A Structural Balance Approach." While a tendency to balance the model positively is anticipated because of the attitude in the communication toward the issue, no predictions can be based on the model for those who have no receiver-to-source attitude. The model allows only for +1 and -l relations, not null or neutral ones. Identifying the source as Republican or neutral should not serve to valence it for those with no strong receiver-to— source attitude, the state assumed for the Ticket Splitters. H2 Repetition in a persuasive political message of a slogan which is ambiguous enough to be broadly interpreted, and serves to link the receiver with the issue of the message, will cause more attitude change than use of the slogan once. Political slogans are typically ambiguous. The one used here is "The door to education is open to your child." An ambiguous slogan about an issue can accomplish several things: First, it can serve to make the issue rele- vant to the voter by suggesting how it affects him. Secondly, its ambiguity makes it easier for the voter to be in agree~ 60,61 ment with it. Wilson has recently engaged in a series of experiments demonstrating that greater agreement is found with ambiguous than with explicit statements. Lazarsfeld gt 1.62 much earlier found a certain degree of ambiguity 60Gary B. Wilson, "Purposeful Ambiguity; An Explora— tory Investigation," Unpublished ditto manuscript, Dept. of Communication, Michigan State University, Spring, 1968. 61Gary B. Wilson, "The Use of Ambiguity as a Message Strategy,“ Unpublished ditto manuscript, Dept. of Communica— tion, Michigan State University, June, 1969. 2Lazarsfeld, The People's Choice. 48 apparently increased the effectiveness of appeals by render- ing them susceptible of various interpretations. Repetition of the legan was operationalized here with some variation. The cover design on all test messages carried a variation of the slogan. Inside the full slogan was used once in half the versions and three times in the others. Klapper63 has summarized a variety of studies which show that repetition with variation serves to remind the receiver of the goal of the persuasion, and simultaneously to appeal to several of his needs and drives. H3 A voter for whom an issue is important will have more positive attitude change toward the issue after read~ ing a persuasive message favorable to the issue than a voter for whom the issue is less important. The voter who conSiders an is-ue important more likely will be one who receives or expects direct or indi— rect benefits if particular action is taken on the issue. In the experimental situation here, the messages tell of educational benefits as the result of action on the issue, and specifically which levels of education receive benefits. Thus it is anticipated that those with family members in schools in Ohio-—and this is used as the operationalization of attaching high importance to the issue-~would have rela~ tively more positive attitude change toward the issue than those who pay the same taxes but receive no direct benefits. 63Joseph Klapper, The Effects of Mass Communication (New York: The Free Press, 1960), pp. 119—120. ?—_ W 49 Canon64 in a further experiment based on dissonance theory has shown that potential usefulness of information has a marked effect on interest in reading either supportive or non-supportive information. 2. Hypotheses About Comprehension of Message Content H4 The more a voter reads of a political issue message, the higher comprehension score he will get, espec- ially if the issue is important to him. A message is a stimulus and here the hypothesis is advanced that some learning will take place in response to this stimulus. The further hypothesis is made that the voter who considers the message topic of importance or utility will be more motivated to learn the facts of the message because the message is more salient to him. H5 A voter will learn facts about a political issue better when he reads a message which asks questions about the issue and answers the questions with facts, than when he reads a message which only presents the facts. 65 . . Berlyne has demonstrated that curiOSity can be intensified simply by putting questions to subjects. Some quests for knowledge start out with an ex- plicit question, either put to the subject by another person or formulated to himself as a consequence of his own thoughts or observations. But this does not always seem to be the case. There are even in- stances of divertive epistemic curiosity, when some— body is eager to learn something new without caring 4Festinger, Conflict, Canon Expt. 65D. E. Berlyne, Conflict, Arousal and Curiosity (New York: McGraw—Hill, 1960), pp. 296—299. SO much what. All specific epistemic behavior must, however, be launched by the equivalent of a ques— tion.66 In an experiment, Berlyne gave an experimental group a prequestionnaire, a list of statements which answered the questions which had been raised, and a postquestionnaire. A control group received only the statements and postques- tionnaire. The experimental group could answer significantly more questions on the postquestionnaire correctly than could the control. Berlyne took this as evidence that questions heighten curiosity, facilitating the retention of facts that answer the questions when they are subsequently encountered. Herrick,67 however, reviewed the programmed learning literature on problem setting questions and found no conclu- sive results demonstrating that questions could affect either rate or amount of learning. He did point up the wide usage of the question—answer format in textbooks, films and film— strips. The hypothesis here is based on Berlyne's findings. The wide usage of the question-answer format makes it of interest to test further. 66Ibid., p. 289. 67Merlyn C. Herrick, "The Effect of Problem—Setting Questions on Rate and Amount of Learning in Programming Teaching Machines," Unpublished (mimeograph) research re— port (Bloomington: Audio Visual Center, Extension Division, Division of Educational Media, School of Education, Indiana University, May, 1962). 51 3. Slogan Recall Hypothesis “6 Repetition of a slogan in a political message will make that slogan better recalled. In this dissertation experiment the interest in using a repeated slogan or a single slogan was to test effect of this message variable on attitude change. Learning theory would suggest, however, that the more a stimulus is repeated the more it would be learned. Thus an appropriate addition to the experiment seemed to be a test of slogan recall to see if the additional repeats increased learning of the slogan. CHAPTER TWO: RESEARCH DESIGN AND METHODOLOGY I. PURPOSE The purpose of the experiment was to test variations of a message about a political candidate or issue on a sample of registered voters drawn from a large enough population to represent a wide spread of socioeconomic characteristics. Interest was in determining: 1. Does attending to political messages have any effect either attitudinally or in comprehension of facts on the entire population of voters? 2. Does attending to political messages have dif- ferent effects, either attitudinally or in comprehension of facts, on various segments of the voting population? 3. Do specific variations of a message containing exactly the same factual content have different effects either attitudinally or in comprehension of facts upon the entire population of voters? 4. Do specific variations of a message containing exactly the same factual content have different effects, either attitudinally or in comprehension of facts, on various segments of the voting population? 52 ——_'—7” 53 II. CHOICE AND OPERATIONALIZATION OF VARIABLES l. The Experimental Tool: Message Content, Channel, Source The decision was made to use facts about a political issue, rather than a candidate, the subject matter for mes— sage content. This avoided the problem of determining rele- vant personality characteristics of a candidate or assessing unmeasurable "charisma." Choice of an issue also made it possible to run the experiment at any time, rather than be- ing limited to a campaign period. Specific criteria were determined for choosing the issue. First, it must affect a broad segment of the voting population. Second, it must be either a new issue upon which recent action had been taken or a long—standing issue upon which recent action had been taken appreciably different from previous actions on the same issue. Thirdly, the is- sue must be one which was identified with a political body to which, or official to whom, a partisan party label could be attached. Finally, the issue must be one about which accurate factual information, and partisan viewpoint infor- mation could be obtained. The decision to use a printed brochure as the chan- nel of communication was made on the basis of (1) cost; ease of delivery to subjects; and (3) greater control. a printed brochure the experimenter could be certain of the exact message a subject had received without any interviewer bias. #—~"W 7 W 7 S4 The choice of source was determined by access to information on a political issue. In the fall of 1969 the Republican State Central and Executive Committee in Ohio was willing to supply information on actions of the 1969 session of the Ohio legislature which had a 64:35 Republican majority in the House; 21:12 Republican majority in the Sen— ate. The experimenter was given access to a summary report of the legislature prepared and issued by the Republican State Central and Executive Committee.68 This presented the partisan viewpoint on actions of the legislature, and emphasized that the major accomplishment of the 1969 session was passage of a $2.277 billion education bill for elementary, secondary and higher education, HB 531. A review of news— paper clippings for the major cities of Ohio on the three days following the close of the session confirmed that HB 531 was considered the highlight of the session. Legislative action on education was selected as mes— sage topic for an additional reason. Interviews of a state— wide sample of voters in Ohio in September, l969, had shown voters considered education one of the major problems facing Ohio. To the open—ended question, "What do you think are the most important problems facing Ohio at the present time?" 68Republican State Central and Executive Committee of Ohio, "Highlights of the 108th General Assembly (First Session)" (Unpublished mimeographed report, Fall, 1969). ————wwi 55 respondents had named a cluster of financial problems as number one, number two was schools need improvement, need better teachers (27.2% of respondents). Next in order were air/water pollution (18.2%) and racial problems/civil rights (15.7%). To the succeeding question, "What do you think is the single most important problem facing Ohio at the present time?" they responded with financial, racial, edu— cation, pollution in that order. Reviewing the action of the 1969 legislative session showed that education was the only one of the first four problems on which significant action had been taken. The importance of education to voters was further demonstrated in the current survey. Prior to reading the experimental message, voters were given a list of issues and asked to rate "how important a problem you think that issue is in Ohio at this time." The rating scale ran from O (extremely unimportant) to 10 (extremely important). The mean rating of education across the entire sample was 8.2. To obtain both a partisan and non—partisan view of the education bill enacted by the legislature the experi— menter interviewed Mr. Howard Collier, Director of Finance, State of Ohio, who gave the viewpoint of the Ohio Executive Branch which was instrumental in developing and backing the bill as part of a thrust toward better state financing of education over the period of the eight year administration of Governor James Rhodes, Republican. An interview with 56 Mr. John H. Hall, Assistant Executive Secretary, Governmental Services, Ohio Education Association, gave perspective on criticisms of the bill as well as approval of some of its provisions. Besides newspaper clippings and editorial com- ment, the literature prior and post-passage of the non—parti- san League of Women Voters was read. Mr. Harold Duryee and others on the staff of the Republican State Central and Ex- ecutive Committee, as well as Mr. Collier, answered the ex- perimenter's numerous questions on the many details of the bill. All of this, including a review of Republican past brochures on state aid to education, was used as background for writing a message about the education bill passed by the legislature. The message emphasized new provisions of the bill, as well as its basic provisions. New were: greatly increased appropriations for vocational education to allow 67% more courses, a 50% increase in funds for higher education, aid to non-public schools in the form of paying part of the salaries of lay teachers, a doubling of extra aid to districts with many students on welfare, and increased classes for the handicapped. Basic, but relatively unchanged, was state support for kindergarten through high school education at a level of about one—third of total costs. The message was composed by the experimenter as if it were a promotional piece for the Ohio legislature. This 57 was accomplished by using some of the actual words and phrases of Mr. Collier and of the Republican report on the legisla- ture. Brochure content was approved as accurate to them by officials in the Republican State Central and Executive Com— mittee. Like most political messages, there was thus no direct attribution of source in the "authorship" sense. The source was instead the source of action on the issue, i.e., the Ohio Legislature. One variable manipulated in the experiment was to give the source a valence or not. To give it a valence, in half the messages, the source was identified with a Political Party Label as Republican legis- lature or Republican-controlled legislature. In the other half of the messages the source was given no valence. The Neutral Label was to call it the Ohio legislature or the state legislature. Feather has demonstrated that a receiver tends to see a source as having the same attitude toward the issue as the communication he presents and as agreeing with the communication he presents.69 To satisfy an Ohio law which requires that political literature carry the name and address of an individual re- sponsible, a very finely printed return address with the dummy name Legislative Support for Schools Committee and the name and address of an Ohio resident was carried on the 69Feather, "A Structural Balance Approach," p. 119. 58 back of the brochure. This was not noticed by receivers during the reading and interviewing portion of the experi- ment. To make the reading level as easy as possible, con- sidering the complexity of the bill which was the subject matter, the final version was edited down twice, once by Dr. Verling Troldahl, Department of Communication, Michigan State University, and once again by the experimenter, both of whom have had journalistic training and experience. This editing involved substitutions of simpler words and shorten- ing of sentences. The final result was a message of 687 words, blocked in eight sections. It had a Flesch Diffi— culty Score of 3.3370 which rates as "Standard." Standard score gives a potential audience of anybody who has completed 7th to 8th grade and includes 75% of U.S. adults. By Flesch’s revised formula7l its Reading Ease score was 52. According to the average sentence length the average of 16.4 words made it easier than "Standard," but the average syllables per 100 words, at 162.5 pulled it up to the next higher category, "Fairly Difficult." The syllable count was large— 1y attributable to such words as "educational," "vocational," 7O Rudolph Flesch, The Art of Plain Talk (New York: Harper and Bros., 1946), pp. 195—204. 71Rudolph Flesch, "The Formula for Readability," Mass Communications, ed. by Wilbur Schramm (Urbana: Uni- versity of Illinois Press, 1960), pp. 419-420. 59 and "legislature," which have high syllable counts although they are fairly common words. Given the issue, they were necessary words to use. For interest, the brochure was designed as if it were a promotional piece dependent for readership upon se- lective exposure by the audience. A commercial graphic artist designed the cover using bold arrows, indicating stimuli which Berlyne has shown serve to arouse attention and curiosity.72 The Gestalt pattern of the design moved the reader towards opening the brochure cover. A headline "In Ohio the door is open" was intended to arouse a need for closure in the reader and be sufficiently ambiguous not to be rejected on the basis of selective exposure because of predispositions about the topic of education. The same cover was used on all eight versions. The criteria for maximum legibility were met: black type on white paper, 10 point type for body, bold headlines.73 2. Message Variables (Manipulated Independent Variables) A large number of message variables which are manip- ulated in actual campaign communications have been identi— fied. The practicable limit on the number which would be manipulated in an experimental situation appeared to the 72Berlyne, Conflict, pp. 63—64. 73Miles A. Tinker, Legibility of Print (Ames: Iowa State University Press, 1963). 60 experimenter to be three. This gave eight (2x2x2) versions of the basic message. The problem was to choose three vari- ables of sufficient interest and variety. Decision was made not to select variables which changed the factual content of the message or its orienta- tion as a positive presentation of the 1969 education bill passed by the Ohio legislature. The variables selected and the manner in which they were operationalized were: a. Political party label In half of the versions (Appendices I, II, III, IV) the legislature was described as the Republican legislature or the Republican-controlled legislature. In the other half (Appendices VI, VI, VII, VIII) of the versions it was neu— trally identified as the Ohio legislature or the state leg- islature. As pointed out earlier, this was to give a val- enced and a non-valenced (control) source. This variable was chosen as one for which voters have some value orienta— tion which might affect their attitudes toward the source and hence toward the message itself. b. Question and answer versus straight descriptive format Berlyne74 has demonstrated that questions preceding an informational message increase learning of the facts in the message. This variable was chosen as one which might 74Berlyne, Conflict, pp. 296—297. 61 affect receiver's comprehension. So as not to alter factual content, the wordings of the messages were identical except that the question-answer version (Appendices I, II, V, VI) had questions before each subject matter item, whereas the straight descriptive version (Appendices III, IV, VII, VIII) had only labeling heads. The answers to the questions were the same wording as the straight descriptive paragraphs. c. Slogan repetition The slogan "The door to education is open to your child," was used once in half the versions (Appendices I, III, V, VII) and repeated for a total of three exposures in the other half (Appendices II, IV, VI, VIII). Addition— ally a variation of it was used on the cover and in the open- ing paragraph on all versions. Slogans—-sufficiently ambigu— ous as to be broadly interpreted-—are so frequently used in campaigns it was a matter of interest to see if the extra repetition could cause attitude change and whether the extra repeats would make the slogan better recalled in a post— exposure test. 3. Audience Variables (Measured Independent Variables and Control Variables) The messages were presented to registered voters. Interest was in seeing what message effects were obtained with the total audience and whether the effects occurred differentially with subsets of the audience. The following were chosen as audience variables to be measured for use as independent variables. Measurement was by the questionnaires 62 shown as Appendix Ix (experimental group) or X (control group) and XI (experimental and control). a. Voting Behavior Type This was operationalized by asking the respondent, "In the last General election in which you voted, which an— swer on this card (HAND CARD) best describes how you voted for state and national offices such as Senator and Congress- man?" The card contained the answer choices: Straight Democratic Mostly Democratic Few more Democrats than Republicans About equally for both parties Few more Republicans than Democrats Mostly Republican Straight Republican Don't know mummbwmw Ahh/‘Kf‘flflfl VVV‘VVVVV Categories 1 and 2 were collapsed into Strong Demo- crat, Categories 3, 4, 5 into Ticket Splitters and 6 and 7 to Strong Republican. Respondents in Category 8 were dis— carded. b. Party Self Identification This was operationalized by asking the voter, "Gen- erally speaking, do you consider yourself a Republican, Democrat, or what?" Answers were coded: Republican, Demo— crat, Independent, Don't Know. Later answers were used only for determining the self-identification of the Ticket Split— ters. c. Importance of Education Issue This was operationalized by asking the respondent, 63 "Is anyone in your immediate family-—children, husband/wife, or yourself in school in Ohio now?" (NOT grandchildren, brothers, sisters). Answers were coded Yes or No. A pre- test interview with nine voters in Toledo in late November, 1969, showed that grandparents did not follow what was hap- pening in the schools or with school financing, but several mentioned that they "used to when my children were in school." Following pretest, the question was narrowed to include only the immediate family. The importance of particular types of education was similarly defined as having someone in the immediate family in school by succeeding filter questions: "Is that public or nonpublic (parochial, Catholic, private) school?" Answers were coded Public and/or Nonpub— lic, with Yeses in Public coded as more importance to public education and Yeses in Nonpublic coded as more importance to nonpublic education. A demographic question, "What is your religion," was also used as another measure of attach— ing more importance to nonpublic education if the respondent answered Roman Catholic. Respondents with family members in school were asked, "What level of school?" to determine the relative importance to them of K—12, vocational and higher education. d. Demographic Variables Audience variables, which were measured for use in comparing the control group and experimental groups, were largely demographic: 64 Age Last grade of school attended Religion Sex Income Last grade of school attended was used as the con— trol variable Education Level, collapsed to two categories: Did not complete high school High school graduate or above 4. Message Effect Variables (Dependent Variables) a. Attitude change toward the educational bill passed by the Ohio legislature Before reading the message brochure, respondents were read four statements about the legislature's efforts to im— prove education and its financing in Ohio. Two statements worded positively were alternated with two worded negatively. Respondents were given a scale card: Strongly Agree-—Agree-- Don't Know--Disagree--Strongly Disagree. The l-5 scores were coded in the direction that a five was most favorable toward the legislative action on education and l was least. The same attitude statements were administered again after reading the brochure. Attitude change was computed as the difference between the total score of the four before and the four after statements. The statements were: 1. Here is the first statement: In general, I think the legislature did a good job in trying to improve educa- tion in Ohio this year. 2. In my opinion, the legislature put the extra money they gave to education this year in the wrong places. 6S 3. I think the legislature did a better job of trying to solve the education problems in Ohio than most earlier legislatures have done. 4. In my opinion, the legislature did NOT improve educa- tional opportunities in Ohio. The six inter-item correlations on the attitude ques- tions administered before reading and to the control group ranged from .219 to .496, and the median was .416. b. Comprehension of facts of the message Comprehension was measured by administering a seven- question multiple choice quiz on facts about the 1969 edu- cation bill which were presented in the message brochure. Possible comprehension scores thus ran O—7. The quiz was administered to a control group who did not read the brochure, and to the experimental group, each of whom read one of the eight versions of the brochure. Comprehension gain result— ing from reading was considered as the difference between means of the experimental and control groups. Comprehension differences from different versions of the brochure were the difference in mean scores of those who read the various experimental versions. c. Attitude toward the message Immediately after reading the message brochure, readers were asked: "What do you think of this brochure?" After their first answer one probe, "Anything else?" was asked. The open ended responses were coded in six cate— gories, with multiple responses allowable. l. Informative, gave facts, learned a lot (these replies concerned only the message brochure). -_+'—— 66 2. Good, liked, positive, believed, true, agrees (these replies concerned both the message bro- chure and the education bill action). 3. Negative, disliked, disagrees about education. 4. Questions truth and accuracy. 5. Political prOpaganda, Republican, one-sided, slanted, public relations piece, advertisement. 6. Not interested, couldn‘t understand, too much reading, blind. d. Slogan recall Slogan recall was operationalized by giving the re— spondent a list of four slogans and asking him, "Would you check whether you have read or heard any of these slogans about education." Three of the slogans were made up by the experimenter and the fourth (third on the list) was "The door to education is open to your child," used in the mes- sage brochure. Answers were coded as to whether the respond— ent recalled the correct slogan only, the correct slogan plus others, or did not recall the correct slogan. III. GENERAL STUDY DESIGN Subjects in the sample were interviewed in their homes using the questionnaires of Appendices IX (experi— mental group) or X (control group) and XI (experimental and control). Audience variables and message effect variables were measured according to the procedures discussed under operationalization. Members of the control group received no message brochure. Members of the eight experimental groups each were asked, midway in the interview of Appendix ——— 67 IX, to read one of the eight versions of the brochure. These brochures were the tool for the message variable manipula- tions. Interviewers were professionals employed by Market Opinion Research, Inc., Detroit. Data were coded for analyses by frequency counts, percentages, means, standard deviations, chi squares, and factorial design analysis of variance. Analyses were made with programs run on the Michigan State University CDC 3600 computer. Coding of questionnaires, computer programming, and interpretation of results was made by the experimenter. IV. SAMPLE The counties within the state of Ohio were strati- fied by whether or not they were a county in a Standard Metropolitan Statistical area. The metropolitan counties were ordered by the number of households they contained, as were also the non-metropolitan counties. A probability—proportionate—to-size sample of voters was drawn. A total of 100 sample points were selected in the state. The number of areas per county fell according to the number of households in a given county. Within each of the counties where the sample points fell, the interviewing areas were selected by systematically listing household figures for cities and townships and ran- domly choosing per probability—proportionate-to-size sampling procedures. After the city or township was chosen, the same fi— 7 68 method was used within the city or township to choose the given areas and blocks to work. Maps of each interviewing area were made showing the starting point, skip interval pattern and route to walk. The starting point and skip interval were chosen randomly from the Random Digit Book to avoid any bias on the inter- viewer's part in selecting households. If no registered voter was home at the selected household, the interviewer was allowed to substitute the house next door on either side. If nobody was at home at any of the three households, he skipped the interval from the original household and con- tinued until his interviewing quota was filled. The person to interview in a household was controlled by an equal split between male and female respondents. Questionnaires for this interview were attached to questionnaires for another study involving 1000 interviews. The questionnaires for the experimental group in this study were attached to those for the first, second and third inter— views in each of the 100 sampling points. The questionnaires for the control group were attached to those for the fourth interview in every other sampling point. After all data were punched, the sample was checked against up-to-date household statistics. The completion rate in the rural counties was significantly less than in the metropolitan areas. Since no hypotheses of this study were concerned with size of community, no weighting of the sample was made. 69 V. DATA COLLECTION The message brochure used to operationalize message composition variables was written and approved in October, 1969, following interviews the experimenter conducted in Columbus, Ohio, October 6, 7, 8. Printing was done in early November. Interviewing, originally scheduled for mid—November, was delayed until winter, 1970, for reasons extraneous to this study. This delay had no effect upon the timeliness of the message content as the education bill described pro- vides support for education in Ohio from 1969—71. Interviewing was done between February 15 and March 3, 1970, by professional interviewers employed by Market Opinion Research, Detroit, Michigan. The interviewing questions came at the end of an interview made by Market Opinion Re— search for a client interested in candidate and issue per- ceptions. Nothing in the earlier interview biased respond- ents toward regarding the interview as representing a parti- san political group. Those in this experiment who received the Republican labeled brochures had received nothing parti— san until that point. Coding and data analysis followed in March and April, 1970. CHAPTER THREE: FINDINGS I. DESCRIPTION OF SAMPLE Table 1 gives descriptive information about the total sample and about those within each of the eight experimental groups and one control group. Also shown are scores for before-reading attitudes toward the issue. There are no statistically significant differences between the eight experimental groups in regard to sex, edu— cation level, age, voting behavior, or having family members in school in Ohio. Chi—Square values are shown in Table 1. Table 2 lists interviews not completed or dropped from the analysis and the reasons for exclusion. II. ATTITUDE CHANGE TOWARD ISSUE Attitude change toward the issue was measured as the sum of four after-reading attitude questions about the educational issue (possible range of scores, 4—20) less the sum of four before—reading attitude questions (range 4-20). All eight versions of the experimental messages pro- duced attitude change favorable toward the issue, as shown in Table 3. A t—test for difference between (1) the mean attitude change of the experimental groups of 1.03 and (2) a mean of 0 (no attitude change) was significant, t=7.l, df=26l, p (.001. 70 71 . a.mm we Hm>02 mo. .ea u we o.au an H0>mn no. .m n we .m ucmunwaqmum+ ,Uwououow uoc Coswum> mmmnwme Eocz How ea moosaucfl z Hmuoaww '- .mumou oumzdm ch Mo manna one so Downed mmabmaum> New ocu do mazoum amucwEHquxm pcm Houucou mcu cm¢3DmL moucouowwan pcmuewfldmflm 0c mums mumce. . . ,,. ;. - .. . - mawma m NH 0 mm m .i a m_ m as k_mm e,as c,ms m.ms o.mH sumac» wearers. mcaoomu muowmm 11' . Ari , - . .- V Hoocum ca H mm o m a am m m. m .m a mm q me a we a u. r m. - mm Ass Maegan m>mz -:.!I;: II;I; III:.-: o.mm 9.0m m an a pm m mm . mm m.mm o.em . ms e-em cmuwarzdmm «.mm Ass, 9 ca m m. . mm m.mm n mm s mm m mm a we eJVN uwnuesaw umiuma o am n,o m am m a. m «m a an m,se m mm m me m.mm H mm rmwuoema AsvsHSmrmm ocwnos IllTlilllll illiliilllllilllli m.oa m es ._mn m as m as a on o .m g mm b. H new a: s so m.ee “as. H.sm a me m we 0 .m ¢,He a ma 0.0m m.me a,ma voice m mm o w mnme e an e.w. H mm m.aa A w» 0 cm o.oq I am mmamw Ass 004 v om “as s.mw m cm a so 0 as n as m.oq 9 cm a we m.mm + nmum mm m.mm o.m m-em s cm H mm o om m mm 0 am 9 gm m.mm H mm m: onQm mas Hm>mq .usnm m.am Ass 0 mo v.me H em m mm m.vm m we a.wm k om . we mamemm m.mq m.mu e.am o.sm m we m as w.me m am m,mm m.me m,mm mam: has xwm ..Hmm mm mm mm em am an om om me 2 uwmemm Ampv . - coamum> Hobos + W| m a o m w m m a Houucou mmmmmmz m4m2 mo ouusom mo kumoa Rummdmm cmmon\ummdmm cmmon ozv me cmoon mo undoEHuQHuume ucoamuum\wm3mcmcmm mcfluo>ll< uouumm :MDWWH Qm<308 MOZ BzmazmmMQZH .m4m4Hm<> Bzmozmmma mom m4m<8 xm ho MHMMA mo muusom mo wmummo Azoq\cOH:. ojmmH mo wocmquQEHllm uouumm AHH< omwm\uwmm ommm\mcoz ommmv Qflcmuopmmmnl< uouomm :BszZOU mo BZMQZMQMDZH .m4m4Hm<> Bzmozmmmo mom m4m<8 Mm ho mHmNA mo muusom mo mwummo Aummamm camoam\ummawm cmmon 02v mm: smooam mo Aw>audauumwa ucmwmuum\um3mcmcom “mm4m :HZQBZOU M0 Bzmozmmma mom mqm<8 NM m0 MHWHQll< uouumm BZMQZMQMQZH 86 analysis was run with education level controlled as a co- variate. Results, shown in Table 12, were very little dif- ferent from those of the analysis of variance without edu— cation level controlled. H is not confirmed since the use of Question-Answer S format did not significantly affect comprehension. IV. SLOGAN RECALL H Repetition of a slogan in a political message will make that slogan better recalled. Recall of the slogan, "The door to education is open to your child," was tested by asking respondents to check any slogans they had read or heard in a list of four. The third slogan in the list was the message slogan; the others were dummies made up by the experimenter. About 44% of re- spondents checked several slogans, the same percentage whether they had received a No Slogan Repeat or Slogan Repeat version. The percentage who checked the correct slogan only is shown in Table 13. Using as correct answers only those who recalled the correct slogan and checked none of the dummy slogans, a Z test has a value of 1.2, not significant at the .05 level. Thus H is rejected on the basis of a rather strict 6 test of recall. 87 mo.m NNN Nm.va Amouumv mHHmU GHQDHB mH. Hm.H «m.m H em.o HHm>mH .USUmv mHMHHm>OU om. HN.H mv.v N mm.m Qum< COHuumumwcH mm. NH. we. H v0. Gum COHuumumDCH Hv. Hm. om.m N Hs.w Q04 COHDUMHQHCH mm. vm. mv.m N mm.m Qm< COHuumumDCH hm. oo.H os.m N Hv.m um< COHpomumucH Vm. mm. Nv.H H Nv.H QU COHuumuwDCH _r|4w. vo.H om.m H mm.m Om COHuomumHCH em. NN. Hm. H Hm. um COHuomumDCH am. am. NH.N N mN.v o< COHuumumucH- mm. vv. N©.H N MN.m U< COHuumumDCH mm. NH. 50. N em.H m4 COHuumumucH mm. vm. vN.H H VN.H Q uouumm mm. Ho. mo. H mo. U uouumm Nm. Ho. vo. H «0. m uouumm HN. mm.H Hw.m N Nw.HH 4 uouumm \mwm m oumjmw coo: Eoomoum mwumsgm Mo Esm COHumHum> Mo muusom mo mwumma Aumwamm cmoon\umoQom cmmon oze om: camon mo ucsoeHbQHuume ucmHmuum\uo3mcmcmm OCHDO>II< nouumm "mmqm4Hm<> HzmozmmeZH MBOU m< Hm>mH ZOHB Bzmozmmmo mom mHm ho mHmMH