THE EFFECTS or ASSBRTION INTENSITY ON THE CONGRUITY MODEL By William J. McEwen A THESIS Submitted to Michigan State University in partial fulfillment of the requirements fbr the degree of MASTER OF ARTS Department of Communication 1967 Accepted by the faculty of the Department of Communication, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the Master of Arts degree. lg 91% W Director of Thesis ACKNOWLEDGMENTS The author wishes to express his great appreciation and thanks to his advisor, Dr. Bradley S. Greenberg for his continued assistance and guidance throughout the writing of this thesis. The author also wishes to thank the other members of his committee, Dr. Erwin Bettinghaus and Dr. Gerald Miller. For their invaluable assistance in the collection of the data, the author‘wishes to express his gratitude to Mr. Walter Combs, Mr. Jack Basehart, Mr. Gary Wilson, and Miss Ann Christiansen. The author also wishes to express his thanks and appreciation to his wife fer constant understanding, encouragement, and occasional needed prodding throughout the course of his graduate studies and eSpecially during the preparation of this thesis. Finally, the author wishes to eXpress Special thanks to Mrs. Shirley Sherman fer the typing of this thesis as well as her valuable assistance in its final preparation. ii TABLE OF CONTENTS CHAPTER I INTRODUCTION . . . . . . . . . . . . II METHODS O O O O O O O O O O O O O O Pretest . . . . . . . . . . . . . Manipulations . . . . . . . . . . Measurement . . . . . . . . . . . Design . . . . . . . . . . . . . Procedures............ I II RESULTS 0 O O O O O 0 O O I O O O 0 Ratings of Message Intensity . . . Message Persuasibility . . . . . . Test of the Hypothesis . . . . . . Iv DISCLJSSION O O O O O O O O O O O O 0 Summary . . . . . . . . . . . . . Decision . . . . . . . . . . . . . Methodological Shortcoming . . . Theoretical Problems . . . . . . . Implications........... Research Extension . . . . . . . . BIBLIOGRAPHY O O O O O O O O O O O O O O 0 iii 11 13 16 18 19 19 23 23 25 29 33 33 3M 3M 35 37 39 HQ Table Table Table Table Table Table Table ‘ Table LIST OF TABLES PretestResults.......... ExperimentalDesign. . . . . . . . . . . . Control Group Ratings of Message Intensity Experimental Group Ratings of Message In tens ity O O O O O O O O O O O 0 Experimental Group Ratings of Source Experimental Group Ratings of Concept Mean Deviation Scores for Sources . Mean Deviation Scores for Concepts iv 24 26 27 28 30 31 LIST OF APPENDICES Page APPENDIX A: FIRST PRETEST QUESTIONNAIRE . . . . #6 APPENDIX B: FIRST PRETEST RESULTS . . . . . . 6“ APPENDIX C: EXPERIMENTAL PRETEST QUESTIONNAIRE 67 APPENDIX D: EXPERIMENTAL MESSAGES . . . . . . . 73 APPENDIX E: EXPERIMENTAL POSTTEST . . . . . . . 82 ABSTRACT THE EFFECTS OF ASSERTION INTENSITY ON THE CONGRUITY MODEL by William J. McEwen Since the congruity principle was originally postulated, a number of researchers have utilized this model of cognitive interaction in investigating the effects of psychological imbalance. The congruity model has typically been applied to the prediction of evaluative attitudinal shift resulting from inconsistent associations among three basic elements. These three elements are: 1) the receiver's existing attitude toward the source of a message, 2) the receiver's existing attitude toward the topic (or concept) of the message, and 3) the nature of the assertion which relates source and topic. A range of evaluative rating of both source and topic has been provided for by the congruity principle (most frequently a range of seven values from highly negative to highly positive). The congruity formulations have, however, neglected to provide such a possibility for variance in the intensity of the assertion or message. According to the congruity principle, a given message may only assume twO values: positive (associative) or negative (disassociative). This seeming oversimplification of the relationship between source and concept gave rise to the present study. It was hypothesized that, in situations where source and topic are of equal intensity, congruity model predictions of attitude change will conform most closely to obtained William J. McEwen evaluative shift when the message is of the same intensity as that of the source and topic. Two general cases were investigated: a) source and message topic both evaluated as lightly positive (+1); b) source and topic both evaluated as highly positive (+3). Within each case, experimental messages expressing different levels of evaluative intensity (highly negative and mildly negative) were constructed. Judgments of assertion intensity were created by the experimental manipulation of verbs and modifiers connoting varying levels of evaluative intensity. Measurement of resultant perceived message intensity con- firmed the success of the manipulations. Highly negative messages were perceived as more negative than mildly negative messages. ASubjects for the experiment were 161 undergraduate students. Each subject completed a pretest designed to tap existing evaluative attitudes toward a source and a topic. After completing an irrelevant intervening task, subjects were then exposed to either a (-3) or a {-1) experimental message linking the source and the topic. Posttest ratings of source and message topic were then obtained. Results confirmed the hypothesis in terms of the topic variable, but failed to confirm the hypothesis in terms of the source variable. Obtained attitudinal shift toward the message topic confermed more closely to that predicted by the congruity model when the intensity of the message was equal to the intensity of the topic. Failure to find parallel results fer the source variable may be due in part to the fact that obtained source ratings fbr the source designated as highly positive were somewhat William J. McEwen lower than the experimental criteria demanded. The findings of the study suggest that message or assertion intensity is a crucial variable which should be taken into account when making predictions according to the congruity model. Additional consequent effects of variance in message intensity are also suggested. CHAPTER I INTRODUCTION In 1955, Osgood and Tannenbaum postulated the principle of congruity and applied it to the prediction of attitude change. In the years after this initial postulation, a considerable amount of research has been under- taken employing their model for cognitive interaction. Tannenbaum (1956; 1965; 1966; 1966) and Fishbein (1963; 1964), for example, have utilized the congruity model in investigating cognitive imbalance as it applies to person perception, generalization of attitude change, and various strategies for the reduction of persuasion. The notion of cognitive consistency was, however, not unique to Osgood and Tannenbaum. Newcomb (1953), Heider (1958), and Festinger (1957), among others, had previously concerned themselves with theories of cognitive balance or consistency. What has distinguished the congruity principle as proposed by Osgood and Tannenbaum has not been, then, their theoretic rationale. Its uniqueness has derived from the fact that their model has provided a precise mathematical prediction of the consequences of psychological imbalance. The congruity principle not only has predicted a general attitudinal shift toward cognitive consistency, it also has rigorously stated both the intensity and direction of that shift for the interacting elements in an incongruous relationship. No other theoretical model has attempted quite so precisely to predict the results of the cognitive interaction of two concepts. In addition, the congruity model has made Specific use of communi- cation interactions. Balance or imbalance, according to this principle, occurs as a function of the connective assertion. Osgood and Tannenbaum's principle was therefore unique among balance models both in its attempt at affording precision to the prediction of attitude change and in the importance it has placed on the Operation of a message in an interacting relationship. The congruity principle, as stated by its authors, is: "Whenever two signs are related by an assertion, the mediating characteristic of each shifts toward congruence [with that characteristic of the other, the magnitude of the shift being inversely proportional to the intensities of the interacting reactions." (Osgood and Tannenbaum; 1955) The authors have proceeded from this general statement of the congruity principle to a Specific prediction of evaluative cognitive shift resulting from situations involving inconsistent associations. Osgood (1957) has preposed that connotative meaning exists generally in three major dimensions: evaluative, potency, and activity. Factor analytic studies cited by the author have confirmed these notions. Osgood has further stated that attitudes, as generally defined, exist along the evaluative dimension of meaning. Attitudes have been defined by Osgood as predispositions toward evaluative responses. Thus, the congruity principle, having concerned itself with attitudinal shift, has made pre- dictions of cognitive shift along the particular dimension of meaning which has been labelled "evaluative". The congruity principle has f0cused on the relationship of three significant variables: 1) the existing evaluative attitude toward the source of a message; 2) the existing evaluative attitude toward the concept or topic which the source is evaluating in the message; and 3) the“nature of the assertion which relates source and concept in the message. The evaluation of both the source and concept variables may take values ranging from highly favorable (score of 7) to highly unfavorable (score of l), as measured by the semantic differential (Osgood; 1957). The assertion, however, according to the Osgood and Tannenbaum model, may only be associative (positive) or disassociative (negative). Thus the ratings of the source and the topic of the message may range in evaluative intensity from one to seven. One may have varied attitudes toward President Johnson as a message source and toward Medicare as a message topic. Any assertion that Johnson might make concerning Medicare, according to the congruity principle, could nevertheless be merely one of liking or disliking. There has been no provision for degrees of intensity of liking or disliking. The model has not distinguished between a message strongly condemning Medicare and a message indicating moderate disapproval of Medicare. Both would be classified as "disassociative" assertions and the amount of evaluative shift resulting tom the messages would be predicted to be equal. What Osgood and Tannenbaum proposed, therefore, was as follows. While the receiver of an evaluative message may evaluate the tOpic of the message along a rangg_of values, seven in their particular model, the source of the message is provided with only twg_alternative possibilities. He may describe or talk about the topic as "good", in which case his resultant communication behavior would be associative, or he may describe the topic as "bad", in which case his resultant communication behavior would be disassociative. The model has not allowed for different degrees of association or disassociation. Thus, there has been no provision in the congruity model for the source to have, or to be perceived as having, as many varying states of attitude toward the concept as the receiver of the message may have. The authors have acknowledged that the intensity of associative or disassociative assertions migh£_vary. They have stated that the assertion "A gives 100% support to B" appears more intense than "A is cordial to B". Their principle nevertheless has not accounted for such possible differences in intensity. The model has ignored this, perhaps as somewhat irrelevant. At any rate, no research to date has investigated: a) whether variations in assertion intensity arg_indeed possible; or b) whether these variations have results other than predicted by the congruity model; or c) whether such variations would then need to be accounted for when making congruity principle predictions. Osgood has developed a method for determining the evaluative intensity of a message, which has been termed "evaluative assertion analysis" (1959). The author has stated that connectors may vary in the intensity or degree to which they associate or disassociate source and concept. The overall intensity of a given message may be considered, according to this method, as a function of the intensities of the various connectors (verbs) and evaluative modifiers (adverbs and adjectives). Thus, "Source A is extremely in favor of topic X, since it is definitely desirable" is seen as more intense than "Source A may be somewhat in favor of topic X, since it is slightly desirable". The intensities of the modifiers ("extremely"; "definitely") and the verb ("is") in the first example are greater than the intensities of the modifiers ("somewhat"; "slightly") and the verb ("may be") in the second example. Thus the first assertion would be rated as stronger, or more intense than the second assertion. Evaluative assertion analysis has not, however, been widely employed. The author has attested to its laboriousness and complexity. Bach message must first be broken down into the smallest possible units (each consisting of source, assertion, and concept). These simple units must then be judged according to their apparent level of evaluative intensity by a number of judges. The total of the individual evaluative intensities of the simple units is then taken as an index of the overall evaluative intensity of the particular message. A message in which the judged ratings of the simple units were all highly intense (+3) would thus contrast in intensity with a message of a similar length in which the judged ratings were all (+2) or (+1). With longer messages, however, the result is a large task for coders and judges. The complex message must be broken down into its smallest possible units, all anouns must be blacked out, and judges must Inake a rating of the evaluative intensity of each unit thereby obtained. Westley et a1. (1963) have attempted a simplification of the Osgood technique. The authors selected from a given message only those evaluative statements which applied directly to a previously selected topic or concept. Each individual statement thus chained was then rated by judges as to whether it was positive or negative. No degrees of "positive" or "negative" were permitted. The total number of positive and negative statements was taken as an index of the evaluative intensity of the message. It was assumed by the authors that three mildly positive statements made about a given concept were the equivalent of one extremely positive statement. This basic assumption is questionable. An entire message consisting of mild endorsements hardly seems the equivalent in expressed evaluative intensity of even as few as one or two highly favorable endorsements. In any case, the Westley method has not been employed with any greater frequency than the Osgood method. Westley has stated that, while his method may have been an improvement on the Osgood method in terms of (amount of labor required, it also confronts the coder with a large task. Neither appears adequate to the task of rapidly and reliably arriving at some overall measure of the evaluative intensity of a message. Development of adequate measurement procedures appears a worthwhile area for future research and will be discussed in Chapter IV. The measurement attempts of Osgood and of Westley have been noted here merely to point out that some attention has been paid to assertion intensity. Neither of these techniques was employed in the present experiment. Since both procedures are lengthy and perhaps questionable, it was decided to experimentally manipulate intensity and leave the unsolved problem of measuring evaluative intensity for future inquiry. In addition, neither the Westley nor the Osgood method has been applied to the prediction of evaluative shift in terms of the congruity principle. While researchers have acknowledged the possibility of varying levels of intensity, and even developed instruments to attest to this possibility, the congruity model has failed to account for such a possibility. The mathematical formulation of the principle has made no allowance for anything other than a dichotomous (either positive or negative) assertion. At least at this original framework, then, the authors have discarded any notion that assertions might vary in amount of expressed intensity. Osgood and Tannenbaum (1957) have stated that stronger assertions (or more intense assertions) should increase the congruity effect. The authors have contended that, since "gives 100% support to" is more intense than "is cordial to", the former should increase the probability of obtaining results consistent with congruity model predictions. This appears to be fallacious, as the following example may indicate. For ease of exPlication, one of the situations which the authors have termed "congruent" would serve. Congruent relationships occur, according to the authors when source and concept are of equal polarization and: a) of opposite directions if the assertion is disassociative, or b) in the same direction if the assertion is associative. All other possible relationships are incongruent, to a degree dependent upon the extent to which ratings of source and concept deviate from these ideal, congruent conditions. An evaluatively rated (+1) source linked associatively (positively) with an evaluatively rated (+1) concept would be such a case of con- gruence. The congruity model would predict no change as a function of this association. The authors, however, have speculated that greater assertion intensity should lead to a greater strain toward congruity. They have thus claimed that there should be even less chance of any_ attitudinal shift occurring if linkage was accomplished by a very intense (+3) assertion than if a less intense (+1) assertion had been used to link the two elements. For example, assume that the Governor of Nebraska is rated as (+1) and federal regulations to control water pollution is also rated as (+1). Osgood and Tannenbaum's Speculation would predict that a statement from this Governor eXpressing extreme favorability toward federal regulations should be perceived as more congruent than one which expressed mild favorability toward the concept. Two cases exist in the above example. In the first instance, a mildly-liked source is extremely in favor of a mildly liked concept. In the second case, a mildly favored source is mildly in favor of a mildly liked concept. Not only does the first case not seem more cognitively consistent than the second, the first case actually appears somewhat in- congruous. It rather appears that people I like should feel the same as I do (that is, with the same_degree of intensity - not with greater or lesser intensity) toward concepts that I like. A positive source should ideally evaluate a positive concept to the same degree as I do. If this source expresses a greater or lesser intensity of liking or disliking than I feel toward a particular concept, a certain amount of incongruity should result. It appears more consistent for the Governor of Nebraska to exhibit mild favorability toward federal pollution regulations than for him to very strongly praise such regulations and advocate and endorse their passage. It should be psychologically inconsistent for a favored source to express a degree of favorability toward a concept that is different from that which the receiver would eXpress. The authors of the congruity model thus have proposed a notion which does not seem intuitively or logically to hold for most cases. DeSpite some amount of theorizing and a limited quantity of empirical measurement of message intensity, no research has yet investigated the consequences of assertion intensity on the congruity principle. This thesis proposes to do so. More Specifically, it is proposed that the strength of an assertion is a crucial variable which must be taken into account when predicting attitude change. The hypothesis proposed is therefore: H1: When source and topic are of equal intensity, evaluative attitudinal shifts predicted by the congruity principle will more likely be obtained when the assertion linking source and concept is of the same intensity than when it differs in intensity. In order to illustrate the hypothesis, the following possible cases are provided as examples. For the sake of simplicity, only mild and extreme cases are shown. 10 Source Assertion Concept Case +3 -3 or +3 +3 I +3 -1 or +1 +3 Case +1 -3 or +3 +1 II +1 -1 or +1 +1 Case -3 -3 or +3 -3 III -3 -l or +1 -3 Case «1 -3 or +3 -1 IV -1 -l or +1 -1 It would appear that the obtained evaluative shifts occurring in cases where a (+3) source was linked with a (-3) assertion to a (+3) concept would be quite similar to cases where a (-3) source was linked with a (+3) assertion to a (-3) concept. For this reason, Cases I and III would be eXpected to be similar and, through analogous reasoning, cases II and IV would be expected to likewise be similar. In regards to cases I and III, the hypothesis would state that the congruity principle will better predict resultant attitude change when the assertion is highly intense (+3) than when the assertion is mildly intense (+1). In cases II and IV, the hypothesis would state that the congruity model will better predict attitudinal shift when the assertion is mildly intense (+1) than when the assertion is highly intense (+3). CHAPTER II METHODS It is possible to study a total of 3H3 different cases or source- assertion-concept linkage, allowing each variable to assume a range of seven values. Since Osgood and Tannenbaum have stated that all in- congruous situations will result in at least some attitude change, it was decided to limit the investigation to cases of incongruity, thus affording a comparison of obtained shift with predicted shift. Two major subsets of incongruous relationships may be examined: 1) a positive source opposed to a positive concept (or a negative source opposed to a negative concept); and 2) a positive source in favor of a negative concept (or a negative source in favor of a positive concept). Since it was assumed that assertion intensity would yield similar results in both cases (in terms of correspondence of obtained and predicted changes), it was decided to investigate only the first relationship. In order to further reduce the complexity of the experiment, it was decided to investigate only mild and extreme cases within this subject. In keeping with the hypothesis, the investigation was limited to situations where source and concept were of equal intensity or polari- zation. The four cases selected for experimentation were as follows: 11 12 Source Assertion Concept I. +3 -3 +3 II. +3 -1 +3 III. +1 -3 +1 IV. +1 -1 +1 Two different source-assertion—concept combinations were in- vestigated in each case, thus resulting in a total of eight experimental message linkages. The use of different sources, assertions and concepts at each level of evaluative rating was undertaken in order to assure less chance of any resultant effects being topic bound. Obtaining the same results across message linkages would thus increase the generalizability of the experimental findings. The independent variable manipulated in this experiment was the degree of evaluative intensity expressed by the Specially constructed messages. The dependent variables were two. Subjects (n=l77) judged the degree of evaluative intensity perceived in the messages. In addition, subjects' change scores (from pretest to posttest) in attitude toward the message source and topic were obtained for subjects in each of the experimental conditions. Eight eXperimental messages were prepared. All messages presented the source as having negative evaluations of the message topic (a disassociative assertion). Swurces and topics were selected on the basis of consensually positive evaluative pretest ratings from a separate group of students. There were eight experimental groups, each of which received 13 one experimental message. It was decided, when making predictions according to the congruity model, to disregard the "correction for incredulity". This is a mathematical correction factor variably applied to cases of incongruity, with the greatest correction to be used in cases of maximum incongruity. The authors of the congruity model first noted this correction factor (Osgood and Tannenbaum; 1955) when obtained evaluative shifts failed to conform to predicted changes. The utility of the correction term may be suspect, and there appears to have been insufficient additional research to confirm the mathematical corrections made post facto in the original study. Since there was, then, some question about the empirical reproducibility of the correction for incredulity, it was decided to make predictions without accounting for incredulity. Pretest In order to find sources and concepts with a mean evaluative rating at or near (+3) and (+1), 16 sources and 16 concepts were originally pretested. The total of 32 concepts were rated on six semantic differential scales which loaded highly on the evaluative dimension (Osgood; 1957). The scales employed were: a) valuable...worth1ess; b) important...un- important; 0) good...bad; d) fair...unfair; e) nice...awful; f) wise... foolish. The sources and concepts were administered in two halves. Each half contained eight sources and eight concepts. The two halves were randomly given to 121 subjects from introductory advertising classes at Michigan State University. All subjects were asked to rate each source and concept on the six evaluative semantic differential scales. All data means may 1“ be found in Appendix B. From this pretest, three sources and four concepts were selected for the experiment. Only one source was found to have a mean evaluative rating near (+3). It was therefore decided to pair this same source in the experimental messages with each of the two concepts which were rated at or near (+3). The obtained results are presented in Table 1. Higher scores indicate greater favorability. The range of possible scores was from one to seven, with a mean rating of seven equivalent to a rating of (+3), and a mean rating of five equivalent to a rating of (+1). None of the sources and concepts within each level of evaluation were Significantly different from each other. All comparisons between sources and concepts rated as approximately (+3) and sources and concepts rated as approximately (+1) resulted in significant differences.* 0n the basis of the pretest results, the following associations were employed in constructing the experimental messages. (+3) Source linked with (+3) Concept_ 1. George Romney linked disassociatively with student voice in university affairs. 2. George Romney linked disassociatively with the Peace Corps. (+1) Source linked with (+1) Concept 3. Jacob Javits linked disassociatively with permitting Communist Speakers to talk on campus. u. William Scranton linked disassociatively with raising the minimum wage. , *All obtained probabilities for these comparisons were less than .005. 15 Table l. Pretest Results SOURCES Mean Scale Rating CONCEPTS Mean Scale Rating_ Romney 6.02 "Student 6.0u Voice in University Affairs" "the Peace 5.96 Corps" Scranton 4.99 "Raising the 5.08 Minimum Wage" Javits 5.10 "Permitting 5.22 Communist Speakers to talk on Campus" 16 Manipulations The independent variable was the evaluative intensity of the message or assertion. Two eXperimental messages were constructed for each of the four basic source-concept linkage combinations listed above. One experimental message was mildly intense (-l) and the other experimental message for each combination was strongly intense (-3). All messages were negative or disassociative in order to create incongruous situations. Within each source-concept linkage, modifiers and verbs were manipulated so as to result in two messages (one mild; one strong). Words changed were those which would indicate the source's evaluative position toward the message topic as either strongly negative or mildly negative. The basis for the experimental manipulations was the work of Osgood (1959). In discussing evaluative assertion analysis, the author has pointed to a number of high intensity (:3) modifiers (very; extremely; absolutely; definitely; positively) and a number of weak intensity (ii) modifiers (slightly; occasionally; somewhat). Osgood has further discussed the distinctions between high intensity connectors (verbs "to be"; "to have" - most unqualified simple verbs) and weak intensity connectors ("many"; "might" - verbs implying a possible or hypothetical relation between source and concept). In addition, research conducted by Dodd and Gerbrick (1960) has reported differences between highly polarized modifiers (very strongly; very much; complete) and mildly polarized modifiers (moderately; somewhat; a little). 17 On the basis of such reported differences in expressed attitudinal intensity in verbs and modifiers, the eight eXperimental messages were prepared. To avoid excessive repetition, synonyms fer these modifiers were also used. The two messages constructed from each source-concept linkage combination were identical, except for the eXperimental manipulation of adverbs and verbs. All messages were written in the manner of typical newspaper reports. This was done to maximize the believability of the messages and therefore rule out as much as possible the chance that the reader would find the source-concept linkage incredulous. Messages contained an average of approximately 18 manipulations (range from 15 to 20). Broken down according to type of manipulation, each message had an average of 4 verb manipulations (range from 2 to 6) and In adverb manipulations (range 12 to 16). Sample manipulations from the eight experimental messages are listed below. Low Intensity (-1) Message High Intensity (-3) Message staunchly opposed somewhat opposed strongly moderately severely mildly repeatedly' occasionally obviously possibly complete partial must might strongly somewhat highly slightly is may be requires asks definitely probably evidently seemingly no little definitely possibly would might 18 The complete text of the eight messages may be found in Appendix D. Measurement Two dependent variables were measured. The first was the per- ceived intensity of the messages. Subjects were asked, after reading the message, to rate how strongly they perceived the message source had felt about the topic of the message. A seven point verbal rating scale (ranging from "very strongly opposed" to "very strongly in favor") determined the degree of attitudinal intensity subjects felt was expressed by the message source. Two control groups were employed to check the success of the intensity manipulations. Subjects in these groups (n=l6) read experi- mental messages of either strong or mild intensity from which all mention of a Specific source was deleted. Control group procedures were as follows: 1. Subjects rated eight concepts and were given an intervening task. All prior ratings were irrelevant to the control group purpose. 2. Control subjects read either a very intense (-3) message (n=7) or a mildly intense (-1) message (n=9). Messages were randomly selected from the eight experimental messages. All reference to the message sources wre blocked out. The control messages were thus essentially either strong or mild attacks on various topics ostensibly from an unidentified S OUI‘CG o 19 3. Subjects rated the message according to how strongly they felt the unknown source was in favor of or opposed to the message topic. The second dependent variable was attitude change toward source and concept. The same six evaluative semantic differential scales which had been employed with the pretest sample were given to the subjects before and after message presentation. Subjects rated both source and concept on each of the six scales. Change in attitude toward source and concept were analyzed separately. Design Subjects were 177 students from an introductory communication course at Michigan State University. 161 subjects were randomly assigned to the eight experimental conditions. The remaining 16 subjects were randomly assigned to the two control conditions. The design employed is listed in Table 2. Procedures Experimental group procedures, sequentially: 1. Each subject rated eight concepts (four sources; four message topics) on the six semantic differential scales. Only two concepts (one source and one topic) were relevant to each subject's particular exPerimental condition. The other six concepts were included in an attempt to reduce possible consistency or carry-over effects on the posttest ratings. 20 After rating the concepts, all subjects were given an intervening task. This task consisted of reading a message on wire-tapping practices which was unrelated to the purposes of the eXperiment and then rating the concept "wire-tapping” on a number of Scales. The intervening task required approxi- mately 20 minutes to complete. Subjects then read one of the eight experimental messages. Messages were presented as having appeared recently in a major newspaper. Subjects were asked to underline the main points in the message, thus motivating careful, thorough reading of the message. Subjects rated the message on the seven point verbal rating scale according to how strongly in favor of or opposed to the message topic they perceived the source to be. Subjects rated source and topic on nine semantic differential scales, six of which were used in the subsequent analysis. The other three scales were included in a masking attempt. The six critical scales were the same as had been used to obtain ratings before reading the message. The order of the scales and the reflection of the scale ends were changed from the prior measurement. Congruity principle predictions for Shift on both source and concept ratings were determined for each subject. The Specific predictions were obtained in accordance with the formula supplied by Osgood and Tannenbaum (1955). 21 Table 2. Experimental Design (A) (+3) Source linked with (+3) Concept (total n = 83) Romney-the Peace Corps Romney-student voice in university affairs . (-3) Intensity n = 19 n = 20 Message (-1) Intensity n = 21 n = 23 Message (B) (+1) Source linked with (+1) Concept (total n = 78) Javits-permitting Scranton-raising the Communist Speakers minimum wage to talk on campus (-3) Intensity n = 18 n = 19 Message (-1) Intensity n = 20 n = 21 Message 22 7. Obtained shifts in ratings of source and concept were compared with predicted changes. CHAPTER III RESULTS The eXperimental results are divided into three major sections: 1) ratings of message intensity; 2) persuasibility of the exPerimental messages; and 3) tests of the hypothesis. The first section reports the results obtained from the measurement of perceived message intensity. The second section reports changes in attitude toward source and concept resultant from the experimental messages. The third section reports comparisons between predicted and obtained attitudinal shift for both source and concept variables. Ratings of Message Intensity_ The manipulation of message intensity was apparently successful. Subjects in both the experimental and the control groups who received the mildly intense (—1) messages rated the messages as less intense than did subjects who received the highly intense (-3) messages. The two control groups differed significantly in their ratings of how strongly the source felt about the message topic. For these two .groups, all mention of a Specific source had been deleted, thus minimizing the possibility that judgments of a source's position were dependent upon prior attitudes toward the source. Subjects rated the intensity of an unknown source's position. The results are reported in Table 3. All reported values are mean perceived intensity ratings. Higher absolute values indicate greater message intensity. 23 24 Table 3. Control GrOUp Ratings of Message Intensity Treatments High Intensity Message Low Intensity Message (n=7) (n=9) -2.1W* —1.33 *p<.01 25 A t-test for uncorrelated means resulted in a significant obtained difference between the two groups.* The high intensity (-3) message was rated as significantly more intense than the low intensity (-1) message. The perceived message intensity ratings for the eight experimental groups showed similar differences. Results are reported in Table it. Again all reported values are mean scale ratings. In three of the four comparisons the (~3) experimental message was judged by subjects to be more intense than the (-1) message.** In the fourth comparison, differences in ratings of message intensity approached significance.*** It made no difference if the source was well known or anonymous, or whether he was evaluated as highly positive or mildly positive. Subjects perceived the source's position to be more strongly negative toward the message topic in the (-3) condition than in the (-1) condition. Message Persuasibility_ Posttest ratings of both source and concept, when compared with pre- test ratings, indicate at least some change in every case in the direction of greater congruity. Thus the experimental messages seem to have provided some impetus toward changing attitudes. The tendency in all cases was for ratings of source and concept to be somewhat lower (that is, less favorable) following exposure to the experimental messages. This is a _general trend which the notions of cognitive imbalance or incongruity would predict. *p (.01 (one tailed) ** probabilities less than .05 (one tailed) ***p (.10 (one tailed) Table u. 26 Experimental Group Ratings of Message Intensity Source-Concept Linkage High Intensity Message Low Intensity Message a. Romney-Peace Corps -2.17 (n = 19) -0.7H (n = 21) b. Romney-Student Voice -l.80 (n = 20) -O.7O (n = 23) c. Javits-Communist Speakers -2.00 (n = 18) -l.2l (n = 20) d. Scranton-Minimum Wage -2.59 (n = 19) -1.80 (n = 21) jPp (.005 **p (.01 ***p (.10 *“*p (.05 Significance Tests Linkage t Values (High vs. Low Intensity) a. 3.u37* b. 2.513** c. 1.'+31*** d. 1.9H8**** 27 Table 5. Experimental Group Ratings of Source (-l) Intensity Condition (-3) Intensity Condition Sources Pretest Posttest Pretest Posttest Romney (linked 5.51 5.0“ 5.“2 5.19 with Peace Corps) Romney (linked 5.9a 5.25 5.69 u.so with Student Voice) Javits “.81 “.68 “.79 “.52 Scranton “.92 “.85 5.27 5.00 Significance Tests Correlated t Values (-1) Intensity Condition (-3) Intensity Condition 1.3“* 1.21* *both probabilities less than .20 28 Table 6. Experimental Group Ratings of Concept (-l) Intensity Condition (-3) Intensity Condition Concepts Pretest Posttest Pretest Posttest Peace Corps 6.05 5.51 5.92 5.39 Student Voice 6.“2 5.63 6.6“ 6.10 Communist Speakers 5.30 “.“6 5.53 “.60 Minimum Wage 5.08 “.72 5.““ 5.01 Significance Tests Correlated t Values (-1) Intensity Condition (-3) Intensity Condition 5.66* 3,33** *p