k V LIBRARY Michigan State L University g, PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. DATE DUE DATE DUE DATE DUE MSU Is An Aflirmetive Action/Equal Opportunity Inemmion emana-pj THE ROLE OF COMPARATIVE INFORMATION IN ADVERTISING EVALUATIONS, CANDIDATE EVALUATIONS, VOTING PREFERENCES AND ELECTION INVOLVEMENT BY Bruce E. Pinkleton A DISSERTATION Submitted to Michigan State University qu partial fulfillment for the degree of DOCTOR OF PHILOSOPHY Department of Advertising 1992 ABSTRACT THE ROLE OF COMPARATIVE INFORMATION IN ADVERTISING EVALUATIONS, CANDIDATE EVALUATIONS, VOTING PREFERENCES AND ELECTION INVOLVEMENT BY Bruce E. Pinkleton The purpose of this dissertation is to contribute to a greater breadth of knowledge concerning comparative political advertising effects. In order to accomplish this purpose, an experiment was utilized to examine comparative print advertising in a pretest posttest design. Comparative stimuli contained a manipulation of their perceived negativity as rated by subjects in order to allow hypotheses to be tested across differing levels of negativity. The results of the experiment indicate that noncomparative advertising, limited comparative advertising, and moderate comparative advertising possess significantly more information credibility than complete comparative advertising. Noncomparative advertising also possesses significantly more information credibility than limited comparative advertising. Finally, complete comparative advertising possesses significantly more information utility than limited comparative advertising. Findings also indicate that complete comparative advertising has a significantly more negative impact on targeted candidate evaluations than moderate comparative advertising, limited comparative advertising, and noncomparative advertising. These findings are consistent with research findings in both the negative political advertising and negative information effects literature. Additional analysis indicates that sponsoring candidate evaluations are significantly more positive in the noncomparative and moderate comparative advertising conditions than in the limited comparative advertising condition. Noncomparative advertising also significantly increases sponsoring candidate liking when controlling for the effects of testing. This study also reveals evidence of a weak backlash effect. This effect is most notable in the analysis of sponsoring candidate mean spiritedness. Here, complete comparative advertising significantly increases perceptions of sponsoring candidate mean spiritedness when compared with noncomparative and limited comparative advertising. Despite evidence of a backlash effect, it appears as if comparative advertising does not engender the strong backlash effects often associated with negative attack advertising. Finally, both complete and moderate comparative advertising significantly increase involvement when compared with the decreases in involvement occurring in the noncomparative advertising condition and the control group. While this finding is preliminary, it is the first indication that negative political advertising can increase election involvement rather than decrease election involvement. This dissertation is dedicated to my family for their loving support and sacrifice. iv ACKNOWLEDGEMENTS The author wishes to thank Charles Atkin for his insight, direction, and most importantly, patience. The author also wishes to thank Keith Adler, Richard Cole, Gina Garramone, and E. Lincoln James, for their constructive criticism, encouragement, and support. TABLE OF CONTENTS page LIST OF TABLESOIOIOOOOOOOOOO ..... .00... ....... ...Oix CHAPTER ONE: INTRODUCTION........................1 CHAPTER TWO: LITERATURE REVIEW......... ....... ...7 Negative Information Effects.. ....... ........8 Negative Weighting......................9 Impression Alteration...................15 Resistance to Change....................16 Memorability............................18 Negative Political Advertising...............19 Qualitative Descriptions and Analyses...19 Contents and Usage......................22 Voter Perceptions and Responses.........26 Perceptual and Behavioral Effects.......34 Comparative Advertising......................4l Comparative Product Advertising.........43 Comparative Political Advertising.......54 Summary of Relevant Findings.................63 Negative Information....................63 Negative Political Advertising..........64 Comparative Advertising.................68 CHAPTER THREE: HYPOTHESES, RESEARCH QUESTIONS MD MODSOOOOO...OOOOOOOOOOOOIOO72 Conceptual and Operational Dimensions of Stimuli......................................72 Complete Comparative Advertising........74 Moderate Comparative Advertising........75 Limited Comparative Advertising.........76 Noncomparative Advertising..............77 Dependent Variables..........................78 Advertising Evaluations.................78 Candidate Evaluations and Liking........80 Candidate Mean Spiritedness.............82 Candidate Voting Preferences............83 Election Involvement....................83 Justification of Hypotheses and Research Questions....................................84 Advertising Evaluations.................85 Candidate Evaluations, Liking, and Backlash................................89 Voting Preferences......................94 vi (HLKPTER THREE, CONTINUED Moderate and Limited Comparative Advertising Differences.................97 Election Involvement....................98 Summary of Hypotheses and Research Questions...............................100 nethOdOOOOOCOOOOOOOOOOO0.0.0.000....000000000102 Subjects and Design.....................102 Stimuli and Pretesting..................104 Procedures..............................109 Dependent Variable Measurement..........112 Analysis of Hypotheses..................114 Specified Contrasts and Type I Error....116 mm FOUR: RESULTSOOOOOOOOOOOOOOOOOOO000......120 Manipulation Check...........................120 Reliability Measurements.....................123 Hypothesis Testing...........................126 Post Hypotheses One and One(a)...............126 Hypotheses Two and Two(a)...............129 Hypotheses Three and Three(a)...........131 Hypotheses Four and Four(a).............l33 Hypotheses Five and Five(a).............135 Hoc Analysis of Hypotheses..............137 Hypotheses One and One(a)...............138 Hypotheses Two and Two(a)...............14o Hypotheses Three and Three(a)...........142 Hypotheses Four and Four(a).............142 Hypotheses Five and Five(a).............l44 ResearCh QuestionSOOOO0.0......IOOOOOOOOOOIOOI46 Research Question One...................l46 Research Question Two...................147 Research Question Three.................150 Research Question Four..................151 Research Question Five..................154 Research Question Six...................155 Research Question Seven.................158 summary Of FindingSOOOO00....0.00.00.00.00000160 Information Credibility.................161 Information Utility.....................162 Aggregate Candidate Evaluations.........163 Candidate Liking........................164 Candidate Mean Spiritedness.............165 Candidate Voting Preferences............166 Moderate and Limited Comparative Advertising Differences.................167 Election Involvement....................168 CHAPTER FIVE: DISCUSSION AND CONCLUSIONS.........17O Discussion and Implications of Research FindingSOOOOOOI...OOOOOOOOOOOO0.0.0.000000000170 HmtheseSOOOOOOOOOOO0.00......0.0.0....170 vii CHAPTER FIVE, CONTINUED Research Questions......................178 Limitations..................................188 Future Research..............................190 Summary of Conclusions.......................192 APPENDICES APPENDIX A. Experimental Stimuli............195 APPENDIX B. Instrument......................199 APPENDIX C: Biographic Profiles.............212 APPENDIX D: Control Essay...................213 APPENDIX E: Contrast Weighting Specifications..................214 LISTOFREFmENCESOOOOOOOOOOOOOOOO0.0...0.0.000000216 viii Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table WGQG 10 11 12 13 14 15 16 17 LIST OF TABLES page Cell sizes for treatment conditions and control group........................104 Negativity ratings of advertising stimliOOOOOOOOOOOOOO00.0.0000...0.0.0.00109 ANOVA for experimental stimuli negativity...............................121 Mean ratings of experimental Stimuli negatiVitYOOOOO...OOOOOOOOOOOOOOOOOOOO0.0122 Factor analysis of advertising evaluation scores........................124 ANOVA for information credibility........127 Contrasts for information credibility....127 ANOVA for information utility............129 Contrasts for information utility........130 ANOVA for targeted candidate evaluationSOOOOOOOOO0......00.0.00000000132 Contrasts for targeted candidate evaluations.................... ..... ....132 ANOVA for targeted candidate liking.....134 Contrasts for targeted candidate likingOOOO0.0.000...0.0.0.00000000000000135 ANOVA for targeted candidate voting preferenceSOOOOOOOOO0.00.00.00.000000000136 Contrasts for targeted candidate voting preferenceSOOOOOO0.0.0.0000...0.00.00.00137 ANOVA for post hoc analysis of information credibility.................139 Mean ratings of information credibility.....OOOOOOOOOOOOOO...0......139 ix Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table Table 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 ANOVA for post hoc analysis of information utilitYOOOOOOO0.00.00.00.000141 ANOVA for post hoc analysis of targeted candidate liking........................143 Change scores for targeted candidate likingOOOOOOOOOOO0.0.000...0.00.00.00.00144 ANOVA for post hoc analysis of targeted candidate voting preferences............145 Change scores for targeted candidate voting preferences......................l46 ANOVA for sponsoring candidate evaluationSOOOOOOOOOOOI0.0.0.0.000000000148 Change scores for sponsoring candidate evaluationSOOIO......OOOOOOOOOOOOOOOCO0.148 ANOVA for sponsoring candidate mean spiritedness......OOOOOOOOOOOOOOOOOOCO..152 Change scores for sponsoring candidate mean spiritedneSSOOOOOOOOOOOOOOOOOOOO0.0153 ANOVA for election involvement..........159 Change scores for election involvement..159 Contrast weights for information credibilitYOOOOOOOOOO0.0.0.0000000000000214 Contrast weights for information utility...‘...OOOOOOOOOOO00.0.0000000000214 Contrast weights for targeted candidate evaluationSOOOOOOOOOOOO00.0.000000000000215 Contrast weights for targeted candidate likingOOOOOOOOOOOO...0.00.00.00.00000000215 Contrast weights for targeted candidate voting preferences......................215 CHAPTER ONE INTRODUCTION Political scientists have conventionally explained the outcome of elections through the distribution of voter loyalties to political parties. The majority of voters have historically identified with one of the two major political parties, maintained their partisan perspectives as they interpreted political events, and voted primarily along party lines (Salmore and Salmore, 1989). The mass media, however, have contributed to a radical transformation of political campaigns and electoral processes (Denton and Woodward, 1990). This transformation includes changes in the messages and methods of communication between candidates and voters, changes in the roles of media institutions and journalists in political campaign coverage, and changes in the flow of available information concerning events and occurrences preceding and subsequent to elections (Joslyn, 1984). One of the most significant changes in the electoral process is the increased use of paid political communication in the form of political advertising. Political advertising has become so important that some social scientists have suggested that contemporary political campaigns are essentially waged through political advertising (Trent and Friedenberg, 1991). The messages that candidates place in 2 paid media generally include positive messages about them- selves, negative messages about their opponents, responses to charges by opponents, and comparisons of the candidates (Salmore and Salmore, 1989). Probably the most common, and controversial, development in political advertising involves the increasing use of negative advertising. A week before the 1988 presidential election, a Newsweek article declared: "Voters are fed up and turned off by a hail of mudslinging. But it’s likely to get worse as both sides step up their attack-man ads (Martz, 1988)." Media attention during the election quickly faded from the positions of candidates on important national issues, to a discussion of candidate advertising strategies and tactics (Colford, 1988). Accordingly, the popular press proclaimed the 1988 presidential campaign to be the most negative in the history of the United States (Denton and Woodward, 1990). The results of a variety of research, however, indicate that negative political advertising has become a quite common form of contemporary political communication. Sabato (1981) suggested that a third of all political spots in the 19705 were negative and that the proportion of negative ads being used was increasing. In the final week of a 1982 congressional campaign, over 75% of respondents in a mid- Michigan survey reported seeing television commercials in which one candidate criticized his opponent (Garramone, 1984). In 1986, two-thirds of respondents in six Southern states reported viewing negative political advertisements (Johnson and Copeland, 1987). Finally, in a 1990 mid- Michigan survey, 92% of respondents recalled seeing televised advertising in which one gubernatorial candidate ‘was criticized by his opposition (Pinkleton and Garramone, 1992). As Johnson-Cartee and Copeland (1991b) note, negative political advertising has become a staple of federal, state, and local campaign strategies. Recent electoral outcomes, however, have contributed to significant changes in political campaigns. Currently, both incumbents and challengers are likely to employ a mix of strategies, particularly in highly competitive contests (Pfau, Kenski, Nitz, and Sorenson, 1990). Increasingly, these campaign strategies include comparative advertising as a means of communicating negative information about a candidate's opponent to voters while avoiding the stigma attached to purely negative "attack" advertising (Salmore and Salmore, 1989). In fact, the results of a recent content analysis of the political advertising used in three national elections reveals that almost half of the advertising used mentioned both candidates, and 22% of the ads contained explicit candidate comparisons (Boiney and Paletz, 1991). Unfortunately, the research literature concerning comparative political advertising is extremely limited. Merritt (1984) and Gronbeck (1985) were among the first social scientists to recognize the role of comparative advertising in political campaigns. Both identified 4 comparative political advertising as one of the methods used by political candidates to communicate negative information to voters about an opponent. Merritt (1984) suggests that comparative advertising is isused by one candidate in order to claim superiority over another candidate. This can be contrasted with the purpose of negative advertising which is to identify a competitor for purposes of imputing inferiority (Merritt, 1984). Gronbeck (1985) identifies three types of negative political advertising including: 1) implicative advertising containing an implication or innuendo regarding an opponent; 2) assaultive or attack advertising containing a direct, personal attack on an opponent; and, 3) comparative advertising containing an explicit comparison between candidates. Gronbeck suggests that comparative advertising focuses on both candidates, basing its effectiveness on the assumption that voters actively compare candidates when making voting decisions. In a recent reconsideration of candidate-sponsored negative political advertising effects, Johnson-Cartee and Copeland (1991a) describe comparative political advertising as advertising that specifically compares one candidate with another, with the competitive edge in the comparison given to the sponsoring candidate. Johnson-Cartee and Copeland (1991a, 1991b) identify two kinds of comparative political advertising. Direct comparative advertising features the sponsoring candidate in a point-by-point comparison with his or her opposition. Implied comparative advertising compares the candidates using implicative comparisons rather than direct comparisons. Johnson-Cartee and Copeland suggest that political advertising researchers have only examined "direct reference" or "assault ads." This operationalization is limited giving no consideration to implicative political advertising and comparative political advertising as suggested by Gronbeck (1985). In order to provide a more complete and accurate examination of negative political advertising effects, Johnson-Cartee and Copeland (1991a) examined negative political advertising effects using direct attack, direct comparison, and implied comparison advertising. The results of their research indicate that direct comparison advertising produces a greater negative change in the targeted candidate’s pretest to posttest evaluation means than does either direct attack or indirect comparison advertising. In addition, direct attack and direct comparison advertising scored significantly lower on candidate characteristic issue appeals than indirect comparison advertising. While these findings represent a significant attempt at examining the effects of comparative political advertising, they are incomplete by themselves. One means of enhancing the research literature concerning comparative advertising effects is to employ a comprehensive operationalization of direct comparative advertising. This allows for a more complete examination of comparative advertising effects when 6 employed in an experimental research setting. With this in mind, the purpose of this dissertation is to contribute to a greater breadth of knowledge concerning comparative political advertising effects. This will be accomplished through the utilization of a comprehensive array of direct comparative political advertising. This array will contain a manipulation of the perceived negativity of comparative advertising stimuli as rated by subjects. In this way, hypotheses concerning comparative political advertising effects are testable across differing levels of negativity. In order to accomplish this purpose, Chapter One has provided an introduction to negative and comparative political advertising along with a justification of this research topic. Chapter Two contains reviews of the research literature concerning negative information effects and both negative and comparative political advertising effects. This chapter also includes an abbreviated review of the comparative product advertising research literature. Research hypotheses and their supporting arguments are discussed in Chapter Three, along with the methods used for testing them. Chapter Four contains the results of the study. Chapter Five contains a discussion of the study’s findings including the implications and conclusions to be drawn from the study, research limitations, and recommendations for a future course of research. CHAPTER TWO LITERATURE REVIEW The traditional limited effects model of communication campaigns, based on research concerning the impact of campaign messages on political decision making by Lazarsfeld, Berelson, and Gaudet (1948), is increasingly less applicable to contemporary political campaigns. In the late 19408, social scientists contended that audience members were resistant to persuasive messages in the mass media because of apathy, attitudinal defensiveness, or cognitive ineptness (Atkin, 1981; Rogers and Storey, 1987). In the contemporary political era, however, political parties have declined in importance. Single-issue political activism and the prominence of the mass media have rapidly increased. Social scientists have become increasingly sophisticated in their use of research (Denton and Woodward, 1990). As these changes have taken place, political campaigns have also changed, increasing in sophistication and effectiveness. Foremost among these changes is an increase in the use of negative campaigning, as previously discussed (Johnson-Cartee and Copeland, 1991b; Salmore and Salmore, 1989). The following review of literature is divided into three sections in order to provide a complete examination of the research findings concerning negative and comparative 8 information in political campaigns. The first section contains a broad review of research concerning the perceptual and behavioral effects of negative information, including a review of research by Lau (1982, 1985) concerning the impact of negative information on political perception and behavior. Following this review is an examination of the research literature concerning the perception and effects of negative political advertising. Finally, the research literature concerning comparative advertising is examined. This review contains a largely summary consideration of comparative product advertising research, with a more detailed consideration of the limited comparative political advertising literature. Negative Information Effects The social environment in which humans exist provides a seemingly infinite array of information which may be utilized when forming impressions and making judgments regarding social stimuli. While physiological and cognitive limitations prevent the utilization of all available information in processing tasks, the collection of mental processes used in perceiving, thinking, and remembering may be given to qualitative biases more often than quantitative restraints (Kellerman, 1984). Some of these qualitative biases are attributable to the role of negative information in information processing tasks, including impression formation and person perception. Social scientists have examined the role of negative 9 information in a variety of information processing settings, including individual impression formation and judgmental decision making. The results of this body of research indicate that across different events, settings, and persons, negative experiences or negative aspects of stimuli tend to be more influential in information processing tasks than positive experiences or positive aspects of stimuli (Fiske, 1980; Kellerman, 1984; 1989). Perhaps the most well documented effect of negative information concerns the tendency for individuals to weight negative information more heavily than positive information when forming evaluations of social stimuli. Research also indicates that negative information has a greater ability to alter previously existing judgments than positive information; that impressions based on negative information are more resistant to change than impressions based on positive information: and, that negative information may be more memorable than positive information. Negatile e' tin Social scientists have documented a consistent tendency for negative information to be weighted more heavily than positive information in individual evaluations of social stimuli. Here, comparable positive and negative information--experimentally defined as being equidistant from a psychological neutral point--are not weighted equally in decision making and impression formation. Instead, research findings suggest that subjects typically weight 29 £1. (1' e I; 'C' ,. R" Her For VB. in I) '5’“: VV‘ o I" U.‘ o 10 negative information disproportionately when compared with the weighting given to equidistantly valanced positive information. In one of the first attempts to document the weighting given to negative information, Jordan (1965) observed that data obtained in three separate experiments provided evidence indicating that negative and positive information were not symmetrically opposite in their effects. In his review, Jordan noted that the data sets, collected by different researchers for different purposes, indicated an asymmetric effect between liking and disliking. This asymmetric effect was such that, under comparable conditions, negative information had a stronger impact on attitudes and behavior than did positive information. That same year, Anderson (1965) reported that negative adjectives had a stronger impact on impression formation than did positive adjectives when subjects rated their liking for persons described by sets of personality-trait adjectives. In keeping with these findings, research by Feldman (1966) indicates that negative adjectives possess a stronger modifying capacity than positive adjectives. Feldman examined this capacity by measuring the extent to which negative adjectives pull the evaluation of pairs of adjectives in which they appear, toward themselves. Additional research indicates that, given equidistant favorable, neutral, and unfavorable attributes, negative information tends to be weighted more heavily in subject assessments of individual likability (Hamilton and Huffman, o'av le. 38! Afil UV» 'I'EI I.) l 9‘ .- U. 53 VA ‘A fine Hg, ' .F§ .i. :‘v n 11 1971; Hamilton and Zana, 1972). In a test of impression formation by Hamilton and Zana (1972), for example, subjects were presented with a series of stimulus persons created by combining positive and negative attributes with sets of neutral traits. The favorable and unfavorable attributes were not only equidistant from the neutral point on a scale of likableness, but were opposite in meaning. For each stimulus, subjects wrote a description of their impression of the person and rated the likableness of the person. The results of the experiment indicated that not only was negative information weighted more heavily than positive information in impression formation, but that subjects were much more confident in their evaluative judgments of the negative stimulus persons when compared to the positive stimulus persons. Similar research findings have been obtained by Fiske (1980), Hodges (1974), Levin and Schmidt (1969), Warr and Jackson (1976), and Wyer (1970). It is significant to note that documentation of the disproportionate weighting given to negative information is not necessarily tied to experimental settings. There is evidence indicating that negative information is weighted more heavily than positive information when forming impressions of job applicants (Bolster and Springbett, 1961; Springbett, 1958; Webster, 1964), the police (Jacob, 1971: Walker, Richardson, Williams, Denyer, McGaughey, 1972), the courts (Walker et al., 1972), and bureaucratic governmental agencies (Katz, Gutek, Kahn, and Burton, 1975). 12 There is also research evidence indicating that negative information is weighted more heavily than positive information in political perception and behavior. Specifically, Lau (1982) sought to examine the weighting given to negative information in the formation of impressions, and the weighting given to negative information as a consequence of impressions, in a series of three studies. In his first study, Lau (1982) replicated a 1977 analysis by Kernell. Using data from Michigan's 1974 and 1978 congressional elections and controlling for party identification, Lau examined voter reports of presidential approval or disapproval and voter turnout. Lau hypothesized that voters who disapproved of the president’s job performance would vote in larger numbers than voters who approved of the president's job performance. Underlying Lau's hypothesis was the theory that voters vote against congressional candidates aligned with the president’s party as a means of registering their disapproval of the president's activities. Lau found evidence supportive of his hypothesis suggesting that the consequences of negative evaluations are greater than the consequences of positive evaluations. In his next study, Lau (1982) examined the weighting given to negative information in the formation of political evaluations. Data were examined from the 1968, 1972, and 1980 NES/CPS American National Election Studies, in which a national sample of reSpondents was interviewed before and after each election. Open-ended questions containing the 13 reasons for voting for and against each major candidate were coded. Using regression analysis on positive and negative information indices, Lau found strong evidence for the hypothesis that negative information receives more weight than comparable positive information in the formation of evaluations of presidential candidates. In his final study of the series, Lau (1982) examined data from the 1980 NES/CPS American National Election Studies. These data included a series of new items centered around affects and ascribed personality traits (see Kinder, 1978). These affect and trait measures were collected in a pre-election survey and treated separately. For purposes of analysis, Lau standardized the affect and trait measures using regressions to analyze first the positive and negative affect measures, and then the positive and negative trait measures. Lau’s analysis indicated that negative affect and negative traits had much more influence than positive affect and positive traits in the formation of evaluations of presidential candidates. Lau took these results to be supportive of his hypothesis. In 1985, Lau again undertook an examination of the role of negative information in political perception. In a continuation of his 1982 analyses, Lau attempted to provide evidence supportive of his earlier findings. In this study, Lau proposed two reasons for greater weighting being given to negative information than positive information in political perception. These are the "figure-ground" and IAH 9v ‘1‘ “at ’5‘ “Li A‘ U'. u... . ’Fa Any F‘ '5 A”. W“ :“a " “A. 14 "cost orientation" hypotheses. According to the figure-ground hypothesis, negative information is made more perceptually salient to voters because it stands out against a largely positive background. Because negative information stands out, it is more easily noticed and processed than positive information. As a test of the figure-ground hypothesis, Lau calculated an overall trust-in-government score by summing scores from individual items in the 1968, 1972, 1974, 1978, and 1980 CPS National Election Studies. The data were then separated into thirds by scores to determine the most trusting third of the sample from the most distrusting third of the sample. The results of regression analysis indicated that, for those most trusting of the government, negative information was weighted much more heavily than positive information in congressional elections supporting the figure-ground hypothesis. According to the cost orientation hypothesis, people are more strongly motivated to avoid costs than to approach gains. Because negative information is linked to the avoidance of costs, negative information has greater motivational sway than positive information. As a test of the cost orientation hypothesis, Lau calculated an overall score for "caring" (i.e., caring about the outcome of a presidential election) by summing individual item scores from the 1968, 1972, and 1980 National Election Surveys. This time, the data were separated into two groups in order to distinguish those who cared most about the outcome of 15 elections from those who cared least about the outcome of elections. The results of the study indicated that an increase in the importance of negative information was greater than the comparable increase in the importance of positive information among those who cared most about the outcome of the presidential elections. This finding is in support of the cost orientation hypothesis. Taken together, this series of studies by Lau (1982, 1985) provides evidence indicating that negative information is weighted more heavily than positive information in political perception and behavior. Impression Alteration Negative information has been shown to exhibit a greater capacity to alter impressions, in addition to being weighted more heavily in perception and behavior. In a test of the role of negative information in impression alteration, Cusumano and Richey (1970) gave subjects incompatible positive and negative information to use in the evaluation of a stranger. Following a diversionary task, narrative descriptions of a low, medium, or high valence difference were administered in a positive-negative or negative-positive order. Subject ratings were made after each block of information was received, and then again, a week later. The final results of the study indicated a disproportionate influence of negative information at all valence intensities. Impressions that were originally 16 negative, but had been revised upward by positive information reverted to a negative standing within a week. Originally positive impressions that had been revised downward remained negative. These findings were replicated by Gray-Little (1973) in a study of Danish subjects. In this study, subjects increased the weight they ascribed to negative information in measurements taken seven to ten days after initial impression measurement. This effect was evident for conditions in which subjects originally received positive information and then received negative information, as well as conditions in which subjects originally received negative information and then received positive information. As an additional note, documentation of the ability of negative information to alter already existing impressions exists outside of experimental research settings. Research concerning the impact of negative information on opinion formation in employment interviews, for example, indicates that negative information has a much greater capacity to alter previously positive evaluations than the ability of positive information to alter previously negative evaluations (Bolster and Springbett, 1961; Springbett, 1958). Resistansetogiange Research evidence indicates that impressions based on negative information are more resistant to change than impressions based on positive information. Two experiments 17 by Briscoe, Woodyard, and Shaw (1967), for example, sought to test the hypothesis that an unfavorable first impression is more resistant to change than a favorable first impression. The design of the first experiment incorporated two orders of information (favorable-unfavorable vs. unfavorable-favorable), two modes of responding, and two control groups. The design of the second experiment was similar to the design of the first, however reversed forms of descriptions were used. The results of both experiments revealed substantial support for the authors' hypothesis, indicating that impressions based on negative information are more resistant to change than impressions based on positive information. It is significant to note that the findings of a number of previously discussed studies are generally supportive of the Briscoe, Woodyard, and Shaw (1967) findings. For example, in the previously discussed experiment conducted by Cusumano and Richey (1970), subject impressions that were originally negative but had been revised upward by positive information, reverted back to their original negative standing within a week. In the previously discussed experiment by Gray-Little (1973) subjects that received negative information and then positive information increased the weight they ascribed to negative information in measurements taken seven to ten days after initial impression measurement. Taken together, these studies provide evidence indicating that impressions based on ne; ‘nv w) u I an i.“ c» U“ {EC 7163 hp . w 18 negative information are more resistant to change than impressions based on positive information. Memorability The results of research concerning attitude formation and negative political advertising effects, to be discussed in greater detail in the following literature review, indicate that negative information may be more memorable than positive information. Specifically, research by Shapiro and Rieger (1989) indicates that the arguments used in negative political advertising may be more memorable than the arguments used in positive political advertising. In the Shapiro and Rieger (1989) study, subjects listened to recordings of six advertisements including positive and negative political spots for a fictitious mayoral/city council election. After performing a diversionary task, subjects were asked memory and attitude questions regarding the political advertising. The results of the study indicated that the negative political advertising was significantly more memorable than the positive political advertising . It should be noted that not all research examining the memorability of negative information is supportive of the Shapiro and Rieger findings. Specifically, research by Beigel (1973) and research by Richey, McClelland, and Shimkunas (1967) failed to find results supportive of the claim that negative information is more memorable than positive information. Thus, the findings of the Shapiro and 19 Rieger (1989) study indicating that negative information may encourage greater memory retention than positive information are still incomplete. Negative Political Advertising The body of research literature available concerning negative political advertising is growing, though still incomplete (Aden, 1988). This literature review is divided into four sections in order to simplify the presentation of research findings and contribute to greater ease of understanding. The first section contains qualitative descriptions and rhetorical analyses of negative political advertising. This information is beneficial as an introduction to the conventions of negative campaigning. The next section contains mostly quantitative descriptions of the contents and usage of negative political advertising. The third section contains research findings concerning voters’ perceptual and behavioral responses to negative political advertising. The final section contains research information concerning the effects of negative political advertising, both within specific contests and across the spectrum of the political process. Qualitafifl Minions and Analxses. While there are several rhetorical/descriptive analyses of political advertising generally (e.g., Benson, 1981; Devlin, 1977, 1982, 1987; Larson, 1982), there are a very small number of studies concerning negative political advertising. The analyses that do exist tend to analyze the ST.- LA. "3 de: the :ef 20 techniques and methods used in negative political advertising, rather than examining the broader effects such advertising has. These analyses are not without merit, however. Despite a history of complaints regarding the use of smear tactics in campaigns, there is little rhetorical or descriptive analysis of negative political advertising until 1971. At that time the Fair Campaign Practices Committee, directed by Samuel J. Archibald, published a report entitled The Pollution of Politio . A substantial portion of the report is given to descriptions of "dirty politics" and campaign case studies, many of them containing a substantial portion of negative advertising. In the report, Archibald (1971) and his colleagues decry the use of unfair political practices. This includes the use of "personal vilification, character defamation...(and) scurrilous attacks on any candidate or his personal or family life;" and, the use of campaign material which "misrepresents, distorts, or otherwise falsifies the facts regarding any candidate" including the use of unfounded accusations aimed at creating or exploiting voter doubts. While it is beyond the scope of this dissertation to provide a detailed discussion of the report's conclusions, .it is interesting to note that the authors express particular concern with the potential abuses of televised gmolitical advertising. In particular, the report suggests 21 that the extensive use of televised political commercials may replace intelligent political debate and lead to a distortion of the issues. Among the few specific conclusions drawn in the report is the observation that sponsors of negative political advertising risk encouraging backlash effects whereby voters sympathize with the targets of negative political ads. Interestingly, this effect has been experimentally supported in research by Garramone (1985). In a more recent attempt to rhetorically analyze negative senatorial advertising, Gronbeck (1985) identified three types of negative ads used by political candidates. The first type of ad is implicative in nature, using innuendo to undermine an opponent rather than a direct attack. Gronbeck suggests that the focus of implicative advertising is the sponsoring candidate. Implicative advertising assumes that voters want information to allow them to determine the candidate that most closely aligns with their own personal views. The second type of negative political advertising Gronbeck identifies is comparative in nature, employing a direct comparison between political candidates. The comparative ad increases its focus to include two or more ¢:ompeting candidates and operates under the assumption that voters actively compare contestants in their political decision making process. The final type of negative advertising identified by Gronbeck is assaultive in nature making use of a direct 22 personal attack on an opponent’s character, motives, associates, or actions. Assaultive advertising focuses negative attention on an opposing candidate, operating under the assumption that voters actually vote against candidates, rather than for them. As is the case with the report of the Fair Campaign Practices Committee, Gronbeck suggests little regarding the actual effects of negative political advertising. MC ten and Usage One content analysis by Wadsworth and Kaid (1987) and two content analyses by Louden (1987, 1990) provide a variety of information concerning the content and use of negative advertising in contemporary political campaigns. Wadsworth and Kaid (1987) content analyzed 805 television campaign commercials to investigate the advertising styles used by presidential incumbents and challengers. The commercials analyzed included nine presidential elections from 1952 to 1984, and included the advertising from the two major party candidates. The ads were coded for a variety of verbal, nonverbal, and production features including format, type of appeals, setting and staging of the commercial, and production techniques. By way of general description, 44% of the ads analyzed ‘were sponsored by incumbents with the remaining 56% of ads Sponsored by challengers. The most popular advertising format was introspective in which the candidate offered jPersonal views on issue positions or personal feelings for 23 both incumbents (23%) and challengers (28%). Incumbents also made significant use of testimonials (21%) and opposition-focused (21%) ads, while challengers made greater use of the question and answer format (17%) followed by opposition-focused ads (16%). Challengers were more likely than incumbents to use a ”call for change" as a means of attacking the record of an opponent. While incumbents were able to remind voters of past accomplishments, challengers tended to point out incumbent failures and advocate changes in administration policies (Wadsworth and Kaid, 1987). Of particular interest for the purposes of this paper, negative opponent-focused ads were slightly more likely to be used by challengers (31%) than by incumbents (27%). In attacking opposing candidates, both incumbents and challengers were more likely to attack the issue stands or consistency of an opponent. The use of humor or ridicule was most often used by candidates in their negative advertising. Incumbents employed this strategy in 21% of their ads and challengers used it in 20% of their ads. In addition, challengers were significantly more likely than incumbents to use negative association as an attack strategy. In his first content analysis, Louden (1987) examined 95 political advertisements from the 1984 North Carolina senate race between incumbent Senator Jesse Helms and Governor Jim Hunt. In his analysis Louden attempted to determine the manner in which candidates appeared in their 24 own political advertising (presence), and the manner in which opponents were portrayed in each other's advertising (negative advertising). It is instructive to note that eighteen months before the election opinion polls placed Hunt comfortably ahead in the race with a 22% lead. By may of 1984, however, Helms enjoyed a 50% to 46% lead and won reelection. Louden (1987) attributes much of the success of the Helms campaign to its use of negative spots. By way of description, approximately 59% of the ads run by Hunt and 67% of the ads run by Helms were negative. Most of the Hunt spots were 30 seconds in length and involved either a policy statement or an attack on Helm's issue stands. Most of Helms' spots also lasted 30 seconds. In addition, Helms made use of a significant number of 10 second negative spots. Much of the Helms advertising attacked the specific issue stands of Hunt with a significantly smaller number of ads containing Helms endorsements. Of particular importance to this dissertation, Louden concludes on the basis of his analysis that negative advertising may be effectively used to undercut an opponent's views while increasing positive voter perceptions of the sponsoring candidate. Regarding the successful use of negative spots, Louden suggests: 1) that negative ads may be successfully employed by political candidates when used to encourage incremental shifts in voter perceptions: 2) that negative spots work when they are in general agreement with voter beliefs; and, 3) that negative 25 advertising does not change voter beliefs as much as it offers alternative perspectives from which to view generally agreed upon "facts." Louden concludes his discussion of negative advertising in the Helms/Hunt race by noting the potential dangers of voter backlash. Louden (1990) examined factors contributing to the efficacy of negative political spot advertising, in a second analysis of advertising in the 1984 Hunt/Helms senate race. In this study, Louden posits that academicians' distinctions between image and issue information in political advertising are false. That is, image and issue ads are not discrete message forms. Instead, the formation of candidates' character information is best achieved through the embedding of image messages within advertising that is issue based (Louden, 1990). Such embedding allows candidates to demonstrate negative information concerning an opponent rather than employing "empty words." This is a more efficient means of establishing consistency of character and helps to explain why the Hunt/Helms advertising is more accurately identified in terms of character assessments rather than specific issues (Louden, 1990). The importance of emphasizing issues is heightened for negative spots, according to Louden, because commercials that are issue based provide a veneer of respectability. This allows them to maintain the perception «among voters that they are fair and informative and helps to avoid backlash effects (Louden, 1990) . 26 The results of Louden's analysis indicate that each candidate's campaigns utilized similar portions of negative to positive advertising. There were three distinct differences between the negative advertising employed by each candidate, however. First, Helms' spots were less complex, employing significantly fewer words than Hunt's commercials. Second, Helms’ ads contained fewer policy issues per spot and most often contained information concerning a single issue. Hunt’s ads, conversely, tended to contain multiple issues, often presented in a "laundry list” format. Finally, Helms' ads were found to be significantly more specific than were Hunt's ads. Louden concludes his analysis with an examination of the existence of each candidate's presence in their negative advertising. The results indicate that Hunt's negative advertising is persistent in visualizing its sponsor. Helms, conversely, never appears in his negative advertising or personally expresses reservations about Hunt. From this analysis, Louden concludes that negative political advertising acts to converge voter decisions on an opponent’s character. The presence of the sponsor in such advertising risks invoking unfavorable evaluations of the Sponsoring candidate's actions (Louden, 1990). Voter mm and Responses Voters are often surveyed in an attempt to understand 'the impact of negative political advertising on voter Perceptions and behavior. In one of the first attempts to 27 explore the effects of negative political advertising, Surlin and Gordon (1977) assessed attitudes toward direct- reference "attacking” political advertising using racial class, socioeconomic status, and value system differences as dependent variables. For purposes of their telephone survey, systematic random samples were drawn from the 1972 metropolitan telephone directories for Atlanta and Philadelphia. In the surveys, items designed to assess perceptions of direct-reference political advertising were constructed using agree-disagree statements. The results of the Surlin and Gordon (1977) survey indicated that lower socioeconomic status respondents perceived political direct-reference advertising to be more unethical, but also more informative than did middle socioeconomic respondents. In addition, blacks found the direct-reference political advertising to be more informative and affective than did whites. Gordon and Surlin (1977) concluded their study with a discussion of the issues surrounding their findings, including the suggestion that future research address the need for a code of conduct related to the use of political advertising. In 1984, Merritt reported the results of a survey examining the impact of negative political advertising in a 1982 campaign for State Assembly Representative in California’s 44th Assembly District. In her study, Merritt 1h3pothesized that negative advertising would produce negative affect toward both the target and sponsor of the advertising; that party identification would influence 28 responses to media coverage and evaluations of candidates; and, that exposure to political advertising would affect voter evaluations of candidates and the criteria for making those evaluations. For purposes of the survey, 314 telephone interviews were conducted with residents of the 44th Assembly District approximately a week before the election. The results of the survey indicated that negative political advertising evoked negative affect toward both the target and the sponsor of the negative advertising. Predispositions toward either candidate in the form of partisan affiliation had a limited affect, however. The survey's results indicated that prior partisan attitudes did not produce selective exposure or retention but did produce selective distortion, tending to influence affective responses toward the advertising. Merritt (1984) interpreted the findings of her study as Supportive of the suggestion that exposure to negative Political advertising may cause negative evaluations of the Sponsor and positive evaluations of the target, particularly al'l-'l<>l"ng constituents who identify with the target's political Party. At the same time, negative advertising may encourage counterarguing and source derogation among party loyalists. Merritt concluded by suggesting caution in the use of negative political advertising noting that party loyalists nay reject negative messages and turn against the sponsor of such advertising . 29 Garramone (1984) attempted to determine voter responses to negative political advertising in a telephone survey of 211 mid-Michigan voters during the final week of the 1982 congressional campaign. In her survey Garramone found that more than 75% of the total sample interviewed reported seeing negative political advertising, with 77% of those having seen such advertising able to recall the ad's sponsor 75% of the survey's or target. Based on what they had seen, respondents expressed disapproval of the negative advertising while 19% expressed approval and the remainder had no opinion . Only 10% of the sample evaluated the ads as "very" truthful, with 54% of respondents evaluating the ads as "somewhat" truthful, and 36% of respondents evaluating the ads as "not at all" truthful. Only 4% of respondents reported feeling more positive toward the campaign as a result of seeing the negative advertising, while 40% reported feeling more negative. Regarding the potential for backlash effects, 10% of respondents reported feeling more positive toward the target of negative political advertising while 15% reported feeling more negative. The remainder of the sample reported that the negative advertising had no influence on their perceptions. Additional findings by Garramone (1984) indicate that of the demographic variables, age was the strongest predictor. Older voters tended to perceive the ads as being less truthful, were less approving of negative political advertising, and tended to be more favorably influenced 30 toward the target of negative advertising than younger voters. Not surprisingly, candidate preference was shown to be a better predictor of perception and impact than demographic variables. Here, voters favoring the sponsor of the negative advertising viewed the ads as being most truthful and were more positively influenced toward the sponsor and more negatively influenced toward the target, than voters favoring the target of negative advertising. Johnson-Cartee and Copeland (1989) conducted two surveys in order to make determinations regarding voter’s conceptions of negative political advertising. In the first survey, citizens of Little Rock, Arkansas were asked two open-ended questions concerning the contents of negative political advertising. The open-ended format allowed respondents to explain their own perceptions of what makes a political ad "fair or acceptable" and "unfair or unacceptable." These responses where then used to create post hoc categories containing 10 types of political advertising content, including: Political record; personal life; issue stands; current or past marriages; criminal activities; family members; voting record; religion; medical llistory; and, sex life. In a second survey, voters in six Southern states were aSked to decide if political advertising containing content fI‘om each of the 10 previously identified categories was fairor unfair. The results of the study indicated a split be‘tween the 10 advertising content categories. Advertising 31 content concerning a candidate's stand on issues, criminal activities, political record, and voting record was rated as being fair by a majority of respondents. Conversely, advertising content concerning a candidate's medical history, personal life, religion, sex life, family members, and current or past marriages was rated as being unfair. A confirmatory factor analysis was used to determine whether the ten items used in the second survey split into categories concerning personal characteristics versus political issues. The 10 items split as predicted with a personal characteristics factor composed of items concerning a candidate’s personal life, marriage, family, religion, medical history, and sex life. A second, political issues, factor emerged containing political record, issue stands, criminal record, and voting record. Perloff and Kinsey (1990) used a mail instrument to survey political consultants in an effort to probe their beliefs concerning the of role political advertising in campaigns. Although their response rate was low (37.5%), over 94% of respondents felt that negative advertising exerts a powerful influence on voters' attitudes, and 74% of respondents agreed that negative advertising intensifies VOters' distrust of politicians. Of perhaps greater interest, over 77% of political consultants believed that voters remember negative il'Iformation better than positive information. Additionally, '75% of respondents suggested that negative ads are more 1ikely to stir people's deepest emotions than positive ads. 32 As an additional note, 64% of respondents disagreed with the statement that "once a negative atmosphere is in place, positive media do not work (Perloff and Kinsey, 1990)." Finally, Pinkleton and Garramone (1992) surveyed 405 voters in mid-Michigan the week preceding the November 1990 elections. Their purpose was to examine the effects of negative political advertising on voter cognition, affect and behavior. A response rate of approximately 70% was obtained. Approximately 92% of those sampled recalled seeing televised advertising in which one gubernatorial candidate was criticized by his opposition, while 81% of respondents recalled seeing televised advertising in which one senatorial candidate was criticized by his opposition. Approximately 72% of respondents found negative political advertising to be not very informative or not at all informative and 87% of those sampled disapproved somewhat or disapproved strongly of the use of negative political advertising. Both positive and negative correlations existed between 'voter characteristics and cognitive and behavioral ‘Variables. Many of these correlations can be generally (Iharacterized as low but significant. Age and the extent to “flaich voters found negative political advertising jInformative was negatively correlated (r=-.17, p <.01). Voting in the past gubernatorial election was negatively ‘Jorrelated with finding negative political advertising informative (r=-.17, p <.01) and approving of the use of 33 negative political advertising (r=-.11, p <.05). Viewing a high number of negative gubernatorial ads correlated negatively with approval (r=-.10, p <.05), while viewing a high number of negative senatorial ads correlated negatively with informativeness (r=-.11, p <.05). Approval of the use of negative political advertising was significantly but weakly correlated with interest in the outcomes of both the gubernatorial (r=.27, p <.01) and senatorial elections (r=.22, p <.01). Finding negative political advertising informative enjoyed a significant, moderate correlation with approval of the use of negative political advertising (r=.47, p <.01), indicating that those who found negative political advertising informative also tended to approve of its use. Informativeness was also correlated with interest in both the gubernatorial (r=.31, p <.01) and senatorial elections (r=.23, p <.01), to a weaker extent. Somewhat surprisingly, the informativeness of negative political advertising was weakly but negatively correlated with the likelihood of voting (r=-.17, p <.01). Multiple regressions.were used to partial variance aiding in the determination of the relative degree of <30ntribution of variables. Both direct and interaction effects proved significant. In terms of likelihood of Voting, total caring about the outcome of the elections igenerated a highly reliable beta coefficient when the effects other variables were partialed out and controlled for (beta=.29, p <.01)- 34 In the same way, an age by informativeness interaction effect also generated a significant beta coefficient when variable contributions were partialed (beta=-.41, p <.05). A plot of the interaction effect between voter age (18-40; 41-91) and informativeness revealed that younger voters were likely to vote no matter how informative they found negative political advertising. Interestingly, older voters who found negative political advertising to be the most informative were less inclined to actually vote than older voters who found negative political advertising less informative. Finally, caring about the outcome of the elections (beta=.25, p <.01); approval of the use of negative political advertising (beta=.20, p <.01); and, finding negative political advertising interesting (beta=.20, p <.01) generated reliable beta coefficients with respect to total interest in the outcome of the elections. Wendflehaflmlm There are several recent experiments examining the perceptual and behavioral effects of negative political radvertising. These studies have examined a variety of ‘tOpics including the roles of sponsor and rebuttal in lmegative political advertising, the appeals and strategies IJsed in negative political advertising, and the effects of ‘Wegative political advertising on the political process. In a»l.985 study, Garramone examined the relationship between negative advertising sponsor and rebuttal factors, and ‘perceptions of the candidates and voting intentions. For 35 purposes of the experiment, communication arts undergraduates at a large Midwestern university were used as subjects. Subjects were randomly assigned to one of three conditions: Candidate sponsor/no rebuttal; independent sponsor/no rebuttal; and, independent sponsor/rebuttal. The results of the experiment revealed that subjects who viewed an independently sponsored negative ad held a more negative View of the target than subjects who viewed the negative ad with a sponsoring candidate. In addition, subjects who viewed an independently sponsored negative ad were less likely to vote for its target. At the same time, these subjects held a more positive view of the target's opponent, and were more likely to vote for this candidate. As an additional note, sponsor by party identification interactions emerged for all dependent variables such that, viewers who did not identify with the sponsor's political party were more sensitive to independent/candidate sponsor effects than were viewers who identified with the sponsoring candidate’s party. Subjects in the rebuttal condition demonstrated a significantly more negative perception of the target's NON > COMP). H1: Moderate and limited comparative advertising will be perceived as more fair and unbiased than noncomparative advertising and complete comparative advertising. Hla: Noncomparative advertising will be perceived as more fair and unbiased than complete comparative advertising. Subject perceptions of the informativeness and usefulness of advertising stimuli are more difficult to predict. These items are reflective of the political issue appeal subscale used by Johnson-Cartee and Copeland (1991a) and should emerge as a single factor in the present study. Unfortunately, existing research evidence concerning the informativeness of political advertising is not entirely clear. As previously mentioned, survey research indicates that voters generally find negative political advertising to be uninformative (Pinkleton and Garramone, 1992). The results of experimental research conducted by Garramone, et al., (1990) however, indicate that negative political advertising may be more informative than positive political advertising, helping to enhance voter image discrimination and facilitating voter decision making. In addition, Hill’s (1989) finding that negative and comparative advertising were rated more highly on a cognitively based global Aad is of some use. This use is limited, however, due to the confounding variables already discussed. An examination of existing product advertising 88 literature provides some additional research findings concerning comparative advertising informativeness. In 1983, Harmon, Razzouk, and Stern conducted a study examining the information content of comparative magazine advertisements. Their research focused on the presence of information cues existing in comparative and noncomparative print advertising. The results of their research indicated that comparative advertising did, in fact, contain more objective information cues than noncomparative advertising. Earl and Pride (1980) examined the role of ad structure, message sidedness, and performance test results on ratings of print advertising informativeness. The results of their study indicated that respondents exposed to comparative advertising rated the ads as being significantly more informative than did respondents exposed to noncomparative advertising. Taken together, these studies provide support, although somewhat sporadic, for the notion that comparative advertising stimuli will be viewed as being more informative than other advertising stimuli. This is most likely to be true for the complete comparative stimuli, since this advertisement contains the greatest amount of comparative material. In the same way, moderate and limited comparative advertising are likely to be viewed as being more informative than noncomparative advertising, as hypothesized (informative and useful: COMP > MOD, LIM > NON). 89 H2: Complete comparative advertising will be perceived as more informative and useful than moderate comparative advertising, limited comparative advertising, and noncomparative advertising. H2a: Moderate comparative advertising and limited comparative advertising will be perceived as more informative and useful than noncomparative advertising. Aggregate evaluations of sponsoring and targeted candidates are initially addressed in this section. Sponsoring and targeted candidate liking is addressed next, followed by discussions of the backlash effect. WW5 As previously discussed, a variety of research exists concerning the impact of negative information, negative political advertising, and comparative political advertising on candidate evaluations. Voters have rated candidates as being less qualified, less honest, less serious, less sincere, less successful, and less of a saver, after exposure to negative political advertising (Kaid and Boydston, 1987). The results of research by Shapiro and Rieger (1989) indicate that the relative impact of negative issue advertising is such that evaluations of the target of the advertising worsen in greater proportion than the evaluations of the sponsor of the advertising. Research findings concerning negative image advertising, however, indicate that subjects evaluate the sponsor no better than they evaluate the target. In the research conducted by Johnson-Cartee and 90 Copeland (1991a), direct comparative advertising produced the greatest significant negative change in a targeted candidate’s pretest to posttest evaluation means. Interestingly, both direct attack advertising and implied comparative advertising also significantly lowered the targeted candidate’s mean evaluation scores. Finally, mention should be made of the negative information effects literature. As previously discussed, the results of research concerning the role of negative information in individual impression formation, person perception, and various other information processing tasks, indicates that negative information tends to be weighted more heavily than positive information when forming evaluations of social stimuli (Anderson, 1965; Feldman, 1966; Jordon, 1965). Research by Lau (1982, 1985) has provided evidence supportive of this effect in political perception and behavior. Based on the above research findings, it is reasonable to expect that complete comparative advertising will produce the greatest negative change in evaluations of the target, due to its negative information content. It is also reasonable to expect that moderate and limited comparative advertising will produce negative change in evaluations of the target to a lessor extent, as hypothesized (target evaluations: COMP > MOD, LIM > NON). 91 H3: Complete comparative advertising will produce greater negative change in evaluations of the targeted candidate than moderate, limited, and noncomparative advertising. H3a: Moderate and limited comparative advertising will produce greater negative change in evaluations of the targeted candidate than noncomparative advertising. Unfortunately, insufficient research evidence exists to make reasonably sound predictions concerning the potentially positive impact of comparative advertising on targeted candidate evaluations. This result would be indicative of a backlash effect and will be discussed in greater detail shortly. In addition, research evidence is lacking concerning the impact of comparative advertising on subject evaluations of sponsoring candidates. These facts lead to research questions 1 and 2. RQ1: Do any of the four types of advertising produce positive change in targeted candidate evaluations? RQ2: Do any of the four types of advertising produce positive or negative change in sponsoring candidate evaluations? Liking Another area worthy of research attention involves the ability of political advertising to alter subjects’ impressions of political candidates. As previously discussed, the results of research concerning the role of negative information in individual impression formation, person perception, and various other information processing tasks, indicate that individuals tend to weight negative information disproportionately when compared with the 92 weighting given to positive information. For purposes of this study, it is of particular importance to note that negative information tends to be weighted more heavily than positive information in assessments of individual likability (Hamilton and Huffman, 1971; Hamilton and Zana, 1972). It should also be noted that subjects are more confident of individual evaluations when based on negative information rather than positive information (Fisk, 1980; Hodges, 1974; Levin and Schmidt, 1969; Warr and Jackson, 1976; Wyer, 1970). Finally, research by Garramone, et al., (1990) indicates that negative political advertising significantly contributes to the polarization of voters’ attitudes toward candidates as measured by expressed candidate liking. Attitude polarization in the experiment was greatest when negative advertising from the sponsor provided a contrast to an opponent’s positive advertising, or lack of advertising. Based on these research findings, it is reasonable to expect that comparative advertising will negatively impact targeted candidate liking, to the extent that such advertising communicates negative information concerning the target to subjects. Complete comparative advertising should have the greatest negative effect on targeted candidate liking, given its negative information content. The same should also be true of moderate and limited comparative advertising to a lessor extent, as hypothesized (targeted candidate liking: coup > MOD, LIM > NON). 93 H4: Complete comparative advertising will produce greater negative change in targeted candidate liking than moderate, limited, and noncomparative advertising. H4a: Moderate and limited comparative advertising will produce greater negative change in targeted candidate liking than noncomparative advertising. Unfortunately, little research evidence is available concerning the impact of comparative advertising on perceptions Of the sponsor. This leads to research question 3. RQ3: DO any of the four types of advertising produce positive change in sponsoring candidate liking? Beetlesh Research evidence indicates that negative political advertising can have a backlash, or boomerang, effect on perceptions Of attacking candidates when voters become more sympathetic toward the target of negative ads, as previously discussed (Garramone, 1985; Merritt, 1984; Shapiro and Rieger, 1989). This is particularly apt to be the case when the negative advertising is image- rather than issue- oriented (Roddy and Garramone, 1988). An examination of the influence Of comparative political advertising on voter backlash leads to inconclusive results, however. Hill’s (1989) examination of comparative, positive, and negative political advertising indicates that voters’ written protocols concerning the sponsor of the advertising were more positive for positive advertising than for negative or comparative advertising. Johnson-Cartee and Copeland (1991a) found no significant 94 evidence Of a backlash effect for direct attack advertising, direct comparative advertising, or implied comparative advertising. The authors suggest that these findings are in keeping with the findings of political consultants rather than the conclusions of academicians. Given these research findings, a question addressing the possible backlash effects resulting from exposure to comparative advertising stimuli is in order. As previously discussed, researchers have examined a number of variables in order to detect the existence of voter backlash, including candidate evaluations and voting preferences. One important aspect Of backlash involves voters’ perceptions of mean spiritedness. When a strong attack on a candidate is perceived by voters as being mean spirited or unjustified, the ad may generate more negative feelings toward the sponsor than toward the target. Research question 4 was developed in order to address this aspect of voter backlash. RQ4: DO any of the four types of advertising produce a positive change in (increase) perceptions of sponsoring candidate mean spiritedness? As an additional note, subjects’ evaluations of candidates, evaluations of candidate liking, and voting preferences will also be examined for evidence of backlash effects. Voting e e es The available research evidence concerning the impact (of political advertising on subject voting preferences is <2haracterized by a lack of consistent findings. From the 95 outset, it should be noted that there is an apparent contradiction in the research findings Of political consultants and academicians. The proprietary research of political consultants has provided evidence indicating that negative political advertising is effective in its intended purpose of influencing the behavior of voters (Johnson- Cartee and Copeland, 1991a). These findings are supportive of political consultants’ positions that negative commercials exert a powerful influence over the attitudes and behavior of voters (Johnson-Cartee and Copeland, 1991b; Perloff and Kinsey, 1990). In contrast, the research efforts of academicians examining political advertising have failed to show that negative political advertising consistently results in its intended impact. Academic research indicates, as previously discussed, that negative political advertising may increase positive perceptions of the target rather than positive perceptions of the sponsor. This is particularly apt to be the case when voters perceive negative advertising to be unfair. In this instance, a backlash effect may result from negative political advertising (Garramone, 1984; Garramone and Smith, 1984). In instances in which a target rebuts the claims made in negative advertising, subjects have demonstrated a significantly lessor likelihood of voting for the target’s opponent than subjects who did not view a rebuttal (Garramone, 1985). The results of research by Roddy and Garramone (1988) 96 indicate that subjects exposed to an issue-oriented negative political commercial expressed a significantly lessor likelihood of voting for the target than subjects exposed to a negative image-oriented commercial. The results of research by Shapiro and Rieger (1989) indicate that when subjects are exposed to a negative image ad, more subjects vote for the target than for the sponsor, indicative of a backlash effect. As a final note, Johnson-Cartee and Copeland (1991a) examined the impact of negative and comparative political advertising on subject voting preferences. An analysis Of the pretest/posttest candidate voting preference scales indicated that the sponsor’s direct attack ad significantly lowered the target’s candidate voting preference score. In the direct and implied comparison advertising conditions, however, the voting preference means did not differ significantly from that of the control group. Despite these findings, it is reasonable to expect that complete comparative advertising will have the greatest negative impact on targeted candidate voting preferences. In this instance, complete comparative advertising contains sufficient negative information to act like direct attack advertising in negatively impacting voting preferences for the targeted candidate. At the same time, the complete comparative stimulus is less likely to engender a backlash effect due to its comparative nature and heightened perceptions of fairness. To a lessor extent, the same should be true of moderate and limited comparative 97 advertising due to their negative content and comparative nature, as hypothesized (voting preferences: COMP > MOD, LIM > NON). H5: Complete comparative advertising will produce greater negative change in targeted candidate voting preferences than moderate, limited, or noncomparative advertising. H5a: Moderate and limited comparative advertising will produce greater negative change in targeted candidate voting preferences than noncomparative advertising. Unfortunately, little research evidence is available concerning the impact of comparative advertising on voting preferences for the sponsoring candidate. This fact contributes to research question 5. RQ5: DO any of the four types of advertising produce positive or negative change in voting preferences for the sponsoring candidate? There is no existing research information concerning the differences between moderate and limited comparative advertising. These stimuli were created for purposes of this experiment, and will be discussed shortly in greater detail. While these operationalizations are generally representative of the different kinds of comparative political advertising used in contemporary campaigns, there is no significant difference between these executions in terms of perceived negativity. The lack of perceived negativity between the stimuli is the primary reason that no differences are hypothesized to exist between moderate and limited comparative advertising 98 in this study. This should not be taken as an absolute indication Of null differences, however. Instead, using moderate and limited stimuli allows for exploratory comparisons to be made between the executions, as presented in research question 6. RQ6: DO moderate and limited comparative advertising differ in their abilities to: -produce positive change in sponsoring candidate evaluations or negative change in targeted candidate evaluations? -produce positive change in sponsoring candidate liking or negative change in targeted candidate liking? -produce positive change in (increase) sponsoring candidate mean spiritedness? -produce positive change in sponsoring candidate voting preferences or negative change in targeted candidate voting preferences? -produce positive change in election involvement? Eleotienlnyelxement There is little research information available concerning the impact of negative and comparative political advertising on election involvement. There is, however, research evidence concerning the influence Of positive political advertising on voter behavior across high and low levels of election involvement. The results of research by Rothschild and Ray (1974) indicate that innocuous political advertising has a strong effect on voting intention in low- level elections (state or local), but not in high-level elections (national). Further, this shift occurs equally among those subjects who are high in election involvement 99 and those subjects who are low in election involvement. In a rare consideration of negative political advertising and election involvement, Garramone (1984) examined the interaction Of involvement with a variety of respondent variables, as previously discussed. The results Of her analysis indicated that, contrary to expectations, voters high in involvement were more likely to be negatively influenced toward the target of negative advertising than voters low in involvement. An examination of this relationship by candidate preference and by education level, however, indicated that this relationship was primarily true for voters favoring neither candidate and for voters high in education. Finally, Garramone, et al., (1990) examined the potential for negative political advertising to increase voter involvement in an election. The authors suggested that negative political advertising may act to increase the election involvement of voters by emphasizing the contest aspect of campaigns. The results of their study, however, indicated that viewing negative political advertising did not significantly increase subject involvement. Research question 7 is in order, given the lack of consistent research findings. RQ7: DO any Of the four types of advertising produce positive change in election involvement? 100 WELWQMWMGS iO_ns The hypotheses and research questions to be examined in this study are summarized here for greater ease of reference. H1: H2: Moderate and limited comparative advertising will be perceived as more fair and unbiased than noncomparative and complete comparative advertising. Hla: Noncomparative advertising will be perceived as more fair and unbiased than complete comparative advertising. (fair/unbiased: MOD, LIM > NON > COMP) Complete comparative advertising will be perceived as more informative and useful than moderate, limited, and noncomparative advertising. H2a: Moderate and limited comparative advertising will be perceived as more informative and useful than noncomparative advertising. (informative/useful: COMP > MOD, LIM > NON) H3: Complete comparative advertising will produce greater negative change in evaluations of the targeted candidate than moderate, limited, and noncomparative advertising. H3a: Moderate and limited comparative advertising will produce greater negative change in evaluations of the targeted candidate than noncomparative advertising. (target evaluations: COMP > MOD, LIM > NON) RQ1: RQ2: H4: Do any of the four types of advertising produce positive change in targeted candidate evaluations? Do any Of the four types of advertising produce positive or negative change in sponsoring candidate evaluations? Complete comparative advertising will produce greater negative change in targeted candidate liking than moderate, limited and noncomparative advertising. RQ3: RQ4: H5: 101 H4a: Moderate and limited comparative advertising will produce greater negative change in targeted candidate liking than noncomparative advertising. (target liking: COMP > MOD, LIM > NON) Do any of the four types of advertising produce positive change in sponsoring candidate liking? Do any of the four types of advertising produce positive change in (increase) perceptions of sponsoring candidate mean spiritedness? Complete comparative advertising will produce greater negative change in targeted candidate voting preferences than moderate, limited, or noncomparative advertising. H5a: Moderate and limited comparative advertising will produce greater negative change in targeted candidate voting preferences than noncomparative advertising. (voting preferences: COMP > MOD, LIM > NON) RQ5: RQ6: RQ7: DO any Of the four types of advertising produce positive or negative change in voting preferences for the sponsoring candidate? Do moderate and limited comparative advertising differ in their abilities to: -produce positive change in sponsoring candidate evaluations or negative change in targeted candidate evaluations? -produce positive change in sponsoring candidate liking or negative change in targeted candidate liking? -produce positive change in (increase) sponsoring candidate mean spiritedness? -produce positive change in sponsoring candidate voting preferences or negative change in targeted candidate voting preferences? -produce positive change in election involvement? DO any Of the four types of advertising produce positive change in election involvement? 102 Method An experiment was conducted in order to test hypotheses and provide answers to research questions. The methods used to conduct the experiment are discussed in the following sections. They include subjects and design, stimuli and pretesting, experimental procedures, and dependent variable measures. This section concludes with a discussion of issues related to data analysis. Melanesia Subjects in the experiment were undergraduate students at a large university in the Southwest. Subjects were primarily juniors and sophomores majoring in communications related fields including public relations, advertising, telecommunications, and journalism. A smaller number of subjects were in business related majors including marketing, accounting, management, and management information systems. All subjects participating in the study were enrolled in sophomore- or junior-level mass communications courses and received extra course credit for participating in the study. Approval of all experimental procedures was Obtained from the Committee for the Protection of Human Subjects prior to the collection of data. Care was taken to select subjects in courses that did not include lectures or readings pertaining to political campaigns or political advertising. In addition, principle and introductory courses were used to provide the majority 103 of subjects for the study to help ensure that subjects would be unsophisticated regarding mass communication theory, consumer behavior, and other subject matter pertaining to techniques of interpersonal and mass persuasion. All subjects were of legal voting age and the sample was approximately equally divided between females and males. Subjects were randomly assigned to one Of four treatment conditions or the control group in a 1 x 5 pretest-posttest design. The subjects in each condition were exposed to different stimuli concerning fictional candidates for a state Senate seat in Idaho. The campaign was presented as an election occurring outside of subjects’ home states in order to make the hypothetical election appear realistic. In each condition and the control, subjects read brief biographic sketches of the competing candidates and completed a set of scales before being exposed to experimental stimuli or the neutral control essay. A total of 228 subjects participated in the study. Twenty instruments were thrown out due to contamination resulting from participation in a pretest or previous participation in the study through another course. One additional instrument was not included in data analysis due to the failure Of the subject to complete a substantial portion Of the scales. A total Of 207 instruments were included in data analysis. NO treatment condition contained less than 40 subjects and cell size was approximately equal, as indicated in Table 1. 104 Table 1 Cell sizes for treatment conditions and control group Conditm su’ ts complete comparative 40 moderate comparative 40 limited comparative 42 noncomparative 42 control 43 Stimuliandfiretestibg The stimuli used in the experiment consisted of print advertising utilizing either a complete comparative execution, a moderate comparative execution, a limited comparative execution, or a noncomparative execution, as previously discussed. The control group read a neutral essay concerning the geography, natural resources, and economy of Idaho. In addition, subjects were exposed to a brief biography of each candidate. Each of these will now be discussed. 5' 1' E E'] Biographic profiles of approximately 160 words were created for each of the candidates (see Appendix C). The profiles were developed specifically for use in the experiment and were based on similar biographies used in other studies examining political advertising. The information contained in each profile was approximately 105 equivalent and contained general information concerning the candidate’s age, education, employment background and volunteer work. The profiles also contained limited discussions of the candidate’s political accomplishments. The results of a t-test between pretest evaluations Of the sponsoring and targeted candidate indicated that subjects’ initial perceptions Of the candidates were not significantly different based on candidate biographies. Adlertisements Four print advertisements were created for the experiment. Each advertisement utilized a similar execution of a different creative strategy, as previously discussed (see Appendix A). The creation of the stimuli involved several steps and required a total Of six pretests. The initial conceptualization of the stimuli was based upon real-world examples Of comparative political advertising. Issue positions were gleaned from a variety of sources including examples of comparative political advertising, academic research findings (i.e., Hellweg, 1979; Shyles, 1986), discussions of political advertising with research experts, and social issues of the day receiving news coverage. Since the political campaign and candidates being used were hypothetical, information contained in the advertising was selected on the basis Of its universal appeal as being a desirable candidate issue position or quality (see Hellweg, 1979). The stimuli were intended to be significantly different 106 from each other in terms of their perceived negativity, based on the amount Of comparative material each execution contained. The complete comparative stimulus contained the most comparative material (six comparative statements) and was intended to be the most negative. The moderate comparative stimulus contained a moderate amount of comparative material (four comparative statements). The moderate stimulus was intended to be less negative than the complete execution, but more negative than the remaining executions. The limited stimulus contained a limited amount of comparative material (two comparative statements). The limited stimulus was intended to be less negative than the complete and moderate stimuli, but more negative than the noncomparative stimulus. The noncomparative stimulus was not intended to be negative and contained no comparative material. The results of pretest 1 were indicative Of the problems encountered in creating experimental stimuli containing significantly different amounts of negative information in a linear relationship. The negativity ratings of the stimuli revealed that the noncomparative execution was rated more positively than the remaining stimuli on a seven-point bipolar scale. The negativity ratings of the remaining three stimuli, however, indicated that the comparative executions bunched together near the midpoint of the scale, within a single standard deviation of each other. 107 In order to make determinations regarding issue positions to include in additional executions of advertising stimuli, subjects rated the value and plausibility of 33 issue stands as part of the initial pretest. For purposes of the ratings, value was defined as the extent to which a quality or issue stand was important, valuable, or beneficial for a candidate to possess. Plausibility was defined as the extent to which it was believable or realistic to expect that one candidate would possess a certain quality while a competing candidate would not possess that same quality. Based on the results of subject ratings, discussions of advertising executions and the test instrument with pretest subjects, and additional research findings, the stimuli were executed a second time with new issue positions. The results of the second pretest were nearly identical to the results of the first pretest. Following the second pretest, the aggregate group Of issue positions, including additional issue stances, were re-evaluated by a different group of subjects in order to gain further information regarding issue positions to include in advertising stimuli. As before, new comparative information was included in the third execution of experimental stimuli, in an effort to generate greater negative ratings. The results Of the third pretest revealed a significant increase in negative ratings of comparative stimuli. The three comparative executions continued to group together, however. Pretests 4 and 5 were undertaken in an effort to find 108 the correct combination of negative, comparative information required to produce the desired effect with respect to the linearity and significance of negativity ratings. The results of the fifth pretest indicated that the stimuli were near to providing the desired results. The noncomparative execution was significantly more positive than the moderate and limited executions, which were significantly more positive than the complete execution. Although the moderate and limited executions continued to group together, both stimuli were retained for use in the experiment for reasons previously discussed. Small changes were made in each of the stimuli and a final pretest was conducted before use of the stimuli in the study. The results of the final pretest revealed an acceptable negativity rating for each Of the advertisements using a seven-point bipolar scale, with one indicating not at all negative and seven indicating very negative. These results are presented in Table 2. 109 Table 2 Negativity ratings of advertising stimuli cu ' n mean stapdato oeviotion noncomparative 2.07 1.15 advertising limited comparative 3.27 1.53 advertising moderate comparative 3.82 1.63 advertising complete comparative 5.3 1.57 advertising QQDLIQl E§§§¥ The neutral essay used in the control condition was created using encyclopedic information concerning the state of Idaho (see Appendix D). The essay was approximately one double-spaced page in length and required approximately one minute to read. The majority of information contained in the essay concerned the geographic properties of Idaho. A smaller portion Of the essay also addressed the state’s economy. Care was taken not to include any information in the essay which would negatively or positively impact subject responses to independent variables. Procedures The experiment was conducted using a double-exposure design. At the beginning of the experiment, subjects were handed instruments turned upside down. Subjects were instructed to leave the instruments upside down until after receiving further instructions. Once all of the instruments 110 were handed out, subjects were instructed to turn their questionnaires over and read and sign the consent form which made up the first page Of the instrument (see Appendix B). After signing the consent form, subjects were instructed to turn to the candidate profiles on page two and to read them. Subjects were given approximately three minutes to read the candidate profiles. After three minutes elapsed, subjects were instructed to turn to page three and complete the scales through page five. Subjects were told not to turn past page five until instructed to do so. Subjects were given four minutes to complete the scales. After four minutes elapsed, subjects were instructed to turn the page and follow the printed instructions. The instruction page, which simply asked subjects to turn to the next page and read the advertisement (or essay), served two purposes. First, it prevented subjects from seeing through the paper to the experimental stimuli while filling out pretest scales. Second, it allowed instructions to be customized for the control condition in which subjects were instructed to read an essay instead Of an advertisement. Subjects were given two minutes to read the advertisement or essay. Following this exposure, subjects were instructed to turn the page and write the thoughts they had concerning each of the candidates. The thought listing exercise served as a diversion between the first and second exposures to experimental stimuli. Thought listing was selected in an 111 attempt to encourage subjects to cognitively process the information to which they had been exposed in the advertisement. Subjects were given approximately four minutes to complete the thought listing exercise. After four minutes elapsed, subjects were again instructed to turn the page and follow the printed instructions. The double exposure design of the experiment was used to help ensure that subjects cognitively processed the information to which they were exposed. In addition, the second exposure helped to provide a closer approximation Of a natural exposure pattern than would otherwise be obtained in an experimental laboratory setting. In this instance, the second exposure approximated the multiple exposures of voters to political advertisements that Often occur during political campaigns. Subjects were given approximately two minutes to read the advertisement or essay. Finally, subjects were instructed to turn the page and complete the scales. These posttest scales were identical to the pretest scales, with the exception Of items related to advertising evaluations. Subjects were given four minutes to complete the posttest scales. Subjects were debriefed after the instruments were collected. The debriefing contained information relating to the purpose of the study, the various experimental stimuli to which subjects were exposed, and the procedures to be followed in data analysis. 112 erondent Variablo Measorement Dependent variables were measured for advertisements, candidates, candidate liking, voting preferences and involvement in the election (see Appendix B). Only brief descriptions of dependent variable measures will be provided here, since the dependent variables have already been discussed conceptually and operationally in some detail. E! !' . E 1 !° 5 Subjects evaluated each advertisement using seven-point scales with bipolar adjectives as anchors. Advertisements were evaluated on six attributes considered important to both political advertising and comparative product advertising success. These attributes include believable, unbiased, informative, interesting, fair, and useful. The contents of each stimulus were also rated for their perceived negativity, in order to provide a manipulation check. WWW Both the sponsor and target of experimental stimuli were rated on a series of seven attributes. These attributes were selected on the basis of their usefulness in previous studies investigating the impact of political advertising on candidate evaluations. Each characteristic was evaluated utilizing a seven-point scale with bipolar adjectives as anchors. The candidate characteristics evaluated were: Intelligent, sincere, believable, honest, 113 persuasive, concerned, qualified, good, and ethical. QandidateLilsimg Candidate liking was measured for both sponsoring and targeted candidates by utilizing seven-point scales with bipolar adjectives as anchors. Subjects provided a measure of impression formation by indicating how much they liked or disliked each Of the candidates. HeanSnirimiiess The mean spiritedness of both the sponsoring and targeted candidates were measured utilizing seven-point scales with bipolar adjectives as anchors. Subject’s perceptions of candidate mean spiritedness were measured to provide an indication of voter backlash. YetingEreferenge Voting preferences for the sponsoring and targeted candidates were measured utilizing seven-point scales with bipolar adjectives as anchors. Subjects were asked to indicate if they definitely would or definitely would not vote for the sponsoring candidate and the targeted candidate. WW Election involvement was measured utilizing seven-point scales with bipolar adjectives as anchors. Subjects indicated how interesting they found the election, how much they cared who won the election, how stimulating they found the election, and how exciting they found the election. 114 Analysis of Hypotheses It is significant to note that all of the hypotheses proposed in this study allow the use Of preplanned contrasts for purposes of data analysis. Preplanned contrasts increase statistical power by allowing the statistical comparison of specific treatment means, rather than an overall analysis all treatment means (Rosenthal and Rosnow, 1985). By definition, any linear combination of treatment means employing fixed lambda weights is a contrast (or comparison) if the sum of the lambda weights is equal to zero (Snedecor and Cochran, 1967). In this instance, polynomial coefficients used to weight treatment means result in a linear combination, or weighted sum of the means (Keppel, 1973). It should be noted that the pairing of a coefficient with a mean is determined by the particular comparison desired. To qualify as a comparison, at least two of the coefficients must be numbers other than zero and the sum of the coefficients must be equal to zero (Keppel, 1973). A distinction should also be made between linear contrasts using orthogonal polynomials, and a set of orthogonal contrasts. A set of orthogonal contrasts is defined by three criteria. The first criterion is that the coefficients used to weight the means sum to zero. The second criterion is that the product Of the weights for each pair of contrasts sum to zero. The third criterion is that the total number of contrasts be no greater than n-l, where 115 n equals the number of treatment conditions (Collyer and Enns, 1986). When these criteria are met, each comparison provides independent, nonredundant, and uncorrelated pieces of information concerning the results of an experiment (Keppel, 1973). Contrasts that meet these criteria are orthogonal as a set, or group. In the instance of nonorthogonal contrasts, only the first criterion has been met. That is, only the coefficients used to weight the means sum to zero. The second and third criteria are no longer necessary (Collyer and Enns, 1986). It is significant to note that orthogonality is not a requirement of meaningful contrasts (Winer, 1971). In fact, the distinctions between orthogonal and nonorthogonal contrasts are relatively minor. Contrasts that are not mutually orthogonal are likely to be used frequently in the analysis of variance (Guenther, 1964). Contrasts should be determined on the basis of their relevance to the research purpose and not on the basis of their orthogonality or nonorthogonality (Guenther, 1964; Rosenthal and Rosnow, 1985). With this in mind, the statistical methods used for testing hypotheses in this dissertation include contrasts based on the weighting Of treatment means using polynomial coefficients summing to zero. These contrasts meet the criterion required for nonorthogonal contrasts but do not meet the criteria required for orthogonal contrasts. In its simplest form, this experiment contains linear, nonorthogonal contrasts, designed to examine the direction 116 and statistical significance of linear relationships according to the hypotheses presented. amines QQLLtT-‘a ts and we I Error The use of specified contrasts within the umbrella of an omnibus F test presents potential problems concerning the overall alpha reliability of significance tests and Type I error. The danger of Type I error--rejecting a null hypothesis when it is true--increases as the number of specified comparisons increase per experiment (Snedecor and Cochran, 1967). In order to better examine this concern, it is useful to distinguish between the Type 1 error rate for a single comparison versus the Type I error rate for the entire experiment. The per comparison error rate uses a single comparison as the conceptual unit for the error rate (Winer, 1971). If several specified comparisons are evaluated in an experiment with an alpha of .05, then the probability of making a Type I error for each specified contrast would be .05 (Keppel, 1973). This can be contrasted with the error rate for the entire experiment which tests the probability of making one or more Type I errors for a set of specified contrasts (Keppel, 1973). It is significant to note that when a large number of comparisons are conducted, even if these comparisons are specified in advance and supported by the results Of previous research and theory, the possibilities of experimentwise Type I error increase. An important distinction must be made at this point 117 between planned (a priori) contrasts and unplanned (a posteriori) tests. In the instance of unplanned, or post hoc analyses, researchers examine mean differences after an omnibus ANOVA has determined that significant differences exist between treatment means. These mean differences may or may not have been expected before an experiment was conducted, but now have relevance to the purposes of the research based on their significance. In this instance, it is necessary to protect against mean differences that appear significant due to error resulting from random comparisons. The protection against making false claims of significance is Obtained by decreasing the ability Of a statistic to detect real differences between means with a known level of probability after an omnibus ANOVA has determined that significant treatment mean differences exist (Snedecor and Cochran, 1967). This is often accomplished using methods developed by Tukey or Scheffe, for example. It is appropriate in this study to use Scheffe’s test in the examination Of research questions and any additional post hoc analyses that are required. Scheffe’s test is appropriate because it is conservative for the pairwise comparison of means. The Scheffe’s test requires larger mean differences for significance than other analytic methods providing a high degree Of protection against Type I error (Keppel, 1973; Snedecor and Cochran, 1967; Winer, 1971). The levels of protection from Type I error demanded by 118 unplanned contrasts are not necessary for most planned contrasts (Keppel, 1973; Rosenthal and Rosnow, 1985; Snedecor and Cochran, 1967; Winer, 1971). When comparisons are planned, researchers are not examining a simultaneous test of all comparisons, but are examining only those comparisons possessing experimental relevance. These are the comparisons for which the experiment was planned, typically based on theory and/or previous research findings (Keppel, 1973). In fact, planned comparisons are appropriate whether an omnibus F is significant or not (Keppel, 1973; Rosenthal and Rosnow, 1985; Winer, 1971). When an unusually large number of contrasts have been planned, additional protection against Type I error may be provided by use of the Bonferroni approach, the Dunn test, or other methods designed to provide Type I error protection (Keppel, 1973; Rosenthal and Rosnow, 1985). It should be noted, however, that this is clearly a special case of additional protection due to an abnormally high number of planned contrasts. When planned contrasts fall within normal usage patterns, additional protection against Type I error is unnecessarily harsh and burdensome (Keppel, 1973; Rosenthal and Rosnow, 1985; Winer, 1971). As a final note, this is particularly true when planned contrasts fall under the umbrella Of an omnibus F with a reasonable probability of statistical significance (Rosenthal and Rosnow, 1985; Snedecor and Cochran, 1967). In this instance, the extra protection provided by a Scheffe’s test, for example, has been likened to "sending out a rowboat to determine if the 119 oceans are calm enough for an ocean liner (Rosenthal and Rosnow, 1985, pg. 46)." In instances of normal usage, it is best to treat all planned comparisons the same way and evaluate each at the same per comparison error rate (Keppel, 1973). CHAPTER FOUR RESULTS The results of the experiment are presented in several sections. The first section contains a manipulation check for subject’s perceptions Of the negativity of advertising stimuli. The second section contains reliability measures for the scales used in the experiment. This section includes the results of a factor analysis of the advertising evaluation scales and the resulting subscales. Hypotheses are tested in order following the manipulation check and presentation of scale reliabilities. The results of post hoc analysis are presented next in order to address questions raised through hypothesis testing. Finally, research questions are addressed and findings are summarized. Manipulation Check A manipulation check was used to test subjects’ perceptions of the negativity of each advertising stimulus. A seven-point scale was utilized with one indicating not at all negative and seven indicating very negative. It should be noted that no subject was exposed to more than one advertising execution. Therefore, subjects’ ratings reflect their perceptions of the absolute negativity of an advertisement, rather than the relative negativity of an advertisement. As an additional note, subjects in the 120 121 control condition also rated the "negativity" of the neutral essay to which they were exposed. A oneway ANOVA was used to examine subjects’ perceptions of the negativity of experimental stimuli to confirm the success of the manipulation. The results Of analysis Of variance indicate that there are statistically significant differences in subjects’ perceptions of the negativity of the stimuli and essay (E (4, 202) = 71.38, p <.001), as indicated in Table 3. Table 3 ANOVA for experimental stimuli negativity SQHIQQ DE 55 £5 E 9 between groups 4 438.03 109.50 71.38 < .001 within groups 202 309.88 1.53 total 206 747.91 A Scheffe’s multiple comparison test was used in order to examine specific differences in the negative perceptions of experimental stimuli. The Scheffe method provides a conservative measure of protection against Type I error for pairwise comparisons of means (Keppel, 1973). An examination Of means reveals a linear effect with respect to negative perceptions of experimental stimuli, as indicated in Table 4. 122 Table 4 Mean ratings of experimental stimuli negativity Condition Mean Standaro Deviation noncomparative 1.57 .887 control 1.67 .919 limited 3.88* 1.31 comparative moderate 4.02* 1.49 comparative complete 5.32** 1.47 comparative * Significantly different from noncomparative advertising and the control at the .05 level. ** Significantly different from all other experimental stimuli and the control at the .05 level. The result of the Scheffe’s test indicates that the complete comparative advertising stimulus is perceived more negatively than each of the remaining stimuli and the control essay at the .05 level. In addition, both the limited and moderate experimental stimuli are perceived more negatively than the noncomparative and control stimuli, also at the .05 level. It should be noted, as previously discussed, that the differences in negative perceptions between the moderate and limited advertising executions are not statistically significant. This is important because it helps to explain a lack of significant differences that emerge between these stimuli in an examination of hypotheses 123 and research questions. Reliability Measurements Multi-item composite scales were used to measure evaluations of experimental stimuli. Composite scales were also used to provide aggregate evaluations of the fictitious political candidates before and after exposure to experimental stimuli, and to provide comprehensive measures of election involvement before and after exposure to experimental stimuli. Each item in a composite measure consisted of a seven-point semantic differential scale with one indicating less of a characteristic and seven indicating more of a characteristic. Cronbach’s alpha reliability coefficients were used in order to determine the reliability of each scale. The reliability of the advertising evaluation scale proved somewhat problematic. As previously discussed, two subscales were expected to emerge based on the results of research by Johnson-Cartee and Copeland (1991a). In fact, an analysis of the overall reliability of the scale (consisting of believable, unbiased, fair, informative, interesting, and useful) revealed a moderate alpha of .72. Given the moderate nature of this alpha, and in order to determine if subscales existed, advertising evaluation scores were submitted to a principal components factor analysis. Criteria for factor retention were a minimum eigenvalue of 1.0 and a minimum variance accounted for of 10%. Two factors emerged as indicated in Table 5. 124 Table 5 Factor analysis Of advertising evaluation scores Fa O i Fa o 2 believable .5758 .6100 unbiased -.0232 .8101 fair .1241 .8794 informative .7326 .2898 interesting .8324 -.2615 useful .7782 .1621 Factor 1 reflects an information utility dimension. This factor emerged with an eigenvalue of 2.65 explaining 44.2% of the variance. Items that loaded high on this factor included informative, interesting, and useful. In addition, factor 1 contained a borderline loading on believability. Factor 2 reflects an information credibility dimension. This factor emerged with an eigenvalue of 1.51 explaining 25.2% of the variance. Items that loaded high on this factor included fair and unbiased. As before, believability appeared with a (stronger) borderline loading. In order to determine the final scales to be used in advertising evaluations, alpha reliabilities were computed for the two factors, each with and without believability. The reliability of the information utility factor without believability was .71. When believability was added to the information utility scale, the alpha reliability improved to .75. In the same way, the reliability of the information 125 credibility scale improved from .70 without believability to .73 with believability. While none Of these reliabilities are ideal, the highest reliabilities for each subscale--higher than the overall reliability obtained for the single composite scale--are Obtained when believability is included in both subscales. The role of believability in each of the subscales makes sense from a conceptual standpoint, as well. It is reasonable to expect that information that is perceived as fair and unbiased is also likely to be perceived as believable. In the same way, it is reasonable to expect that information that is interesting, useful, and informative will also be believable. The believability measure adds depth Of meaning and increased reliability to both scales. For this reason, the final subscales used to evaluate the advertising stimuli are the information utility subscale consisting of informative, interesting, useful, and believable (alpha = .75); and the information credibility subscale consisting Of fair, unbiased, and believable (alpha = .73). The remainder Of the alpha reliabilities for both pretest and posttest measures were reasonably strong. The alpha reliabilities for the sponsor rating scales were .81 for the pretest measure and .90 for the posttest measure. The alpha reliabilities for the target rating scales were .83 for the pretest measure and .87 for the posttest measure. Finally, the alpha reliabilities for the election involvement scale were .85 for the pretest measure and .93 126 for the posttest measure. Hypothesis Testing Tests of hypotheses are conducted in numerical order, followed by post hoc analyses of data related to hypotheses and tests of numerically ordered research questions. Wmmm Hypothesis 1 proposes that moderate and limited comparative advertising will be perceived as more fair and unbiased than noncomparative and complete comparative advertising. Hypothesis 1a proposes that noncomparative advertising will be perceived as more fair and unbiased than complete comparative advertising (MOD, LIM > NON > COMP). Hypotheses 1 and 1a were examined by specifying orthogonal polynomial contrasts to be tested following the application of an omnibus F test for the combined means of fair, unbiased, and believable (see Appendix E for weighting specifications). These form the information credibility subscale. The results of analysis Of variance suggest the existence of statistically significant mean differences among treatment conditions (E (3, 159) = 10.10, p <.001), as indicated in Table 6. 127 Table 6 ANOVA for information credibility Source DE 55 M§ E 9 between groups 3 388.75 129.58 10.10 <.001 within groups 159 2039.56 12.82 total 162 2428.31 An examination of means using contrast analysis indicates that significant differences exist for specified contrasts, as exhibited in Table 7. Table 7 Contrasts for information credibility Contrast Kaine 8.13.! t DE 2 contrast 1 -2.42 .79 -3.06 159 .99 (mod w/ non) contrast 2 1.90 .80 2.37 159 .009 (mod w/ comp) contrast 3 4.32 .79 5.46 159 <.001 (non w/ comp) contrast 4 -2.56 .78 -3.25 159 .99 (lim w/ non) contrast 5 1.76 .79 2.21 159 .014 (lim w/ comp) The results of contrast analysis indicate a modest level of support for hypothesis 1 and full support for hypothesis 1a. While contrasts 1 and 4 in Table 7 appear to possess extremely low levels of statistical probability, these alphas are due to contrast values in the opposite directions Of those predicted. These findings suggest the 128 need for post hoc analysis. In this instance, an ANOVA used in conjunction with Scheffe’s multiple comparison test will provide a conservative pairwise examination Of the contrast between noncomparative and moderate comparative advertising, and the contrast between limited comparative and moderate comparative advertising (see Keppel, 1973; Winer, 1971). Contrasts 2, 3, and 5 indicate the existence of statistically significant mean differences in the directions predicted. In the instance of these contrasts, moderate, limited, and noncomparative advertising, are perceived as being significantly higher in information credibility than complete comparative advertising (MOD, LIM, NON > COMP). Unfortunately, these contrasts do not provide as much information as might normally be expected due to the presence of mean differences in directions opposite of those predicted. With this in mind, hypothesis 1 enjoys partial support, and hypothesis 1a enjoys full support, based on the results Of contrast analysis. While some linear relationships emerge as predicted, the overall linearity suggested in hypotheses 1 and 1a appears to be violated by noncomparative advertising. Post hoc analyses will provide additional information concerning the dependent and independent variable relationships addressed in hypotheses 1 and 1a, particularly where noncomparative advertising is concerned. 129 Hypotheses Two and Two(a) Hypothesis 2 proposes that complete comparative advertising will be perceived as more informative and useful than moderate, limited, and noncomparative advertising. Hypothesis 2a proposes that moderate and limited comparative advertising will be perceived as more informative and useful than noncomparative advertising (COMP > MOD, LIM > NON). Hypotheses 2 and 2a were tested by specifying orthogonal polynomial contrasts to be tested following the application of an omnibus F test for the combined means of informative, interesting, useful, and believable (see Appendix E for weighting specifications). These form the information utility subscale. The results of analysis Of variance suggest the existence of statistically significant mean differences among treatment conditions (E (3, 159) = 3.24, p = .02), as indicated in Table 8. Table 8 ANOVA for information utility SQEEQQ DE 55 HE E 2 between groups 3 214.64 71.55 3.24 .02 within groups 159 3514.26 22.10 total 162 3728.90 An examination of specified contrasts indicates that a significant mean difference exists for contrast 1, as exhibited in Table 9. 130 Table 9 Contrasts for information utility Qontrast Valne DE 2 DE 2 contrast 1 2.11 1.04 2.02 159 .02 (comp w/ lim) contrast 2 -1.04 1.04 —1.00 159 .84 (comp w/ non) contrast 3 -3.15 1.03 -3.06 159 .99 (lim w/ non) contrast 4 .52 1.05 .41 159 .30 (comp w/ mod) contrast 5 -1.57 1.03 -1.51 159 .93 (mod w/ non) In general, statistical support for hypothesis 2 is weak and there is no statistical support for hypothesis 2a. Contrast 1 possesses a statistically significant alpha and the relationship is in the direction predicted. Here, complete comparative advertising is perceived as containing significantly more information utility than limited comparative advertising. Contrasts 2, 3, and 5 possess extremely low levels Of statistical probability. As before, these alphas are due to contrast values in the opposite directions of those predicted. Contrast 4 is the remaining contrast. While the direction of this contrast is consistent with hypothesis 2, it possess a very low level Of statistical probability. The results of these contrasts indicate weak support for hypothesis 2 and no support for hypothesis 2a. In the instance of contrast 1, complete comparative advertising 131 does contain significantly more perceived information utility than limited comparative advertising. The remaining contrasts are either in a direction Opposite of that predicted, or simply do not enjoy a significant level Of statistical support. Post hoc analysis may provide additional information concerning the independent and dependent variable relationships addressed by hypotheses 2 and 2a. Wingeandihreem Hypothesis 3 proposes that complete comparative advertising will produce greater negative change in evaluations Of the targeted candidate than moderate, limited, and noncomparative advertising. Hypothesis 3a proposes that moderate and limited comparative advertising will produce greater negative change in evaluations of the targeted candidate than noncomparative advertising (COMP > MOD, LIM > NON). In order to assess change in evaluations of the targeted candidate, evaluation scales were summed and the pretest mean was subtracted from the posttest mean to create change scores. The differences between treatment change scores were tested first using an omnibus F test. The results of analysis of variance indicate the existence of statistically significant mean differences among treatment conditions (E (3, 157) = 5.79, p <.001), as indicated in Table 10. 132 Table 10 ANOVA for targeted candidate evaluations Sontce DE DE HE E 2 between groups 3 946.49 315.50 5.79 <.001 within groups 157 8554.26 54.48 total 160 9500.76 Following the ANOVA, hypotheses 3 and 3a were tested by specifying orthogonal polynomial contrasts (see Appendix E for weighting specifications). An examination of change score contrasts indicates that significant differences exist, as presented in Table 11. Table 11 Contrasts for targeted candidate evaluations Contrasts yalue DE t DE P contrast 1 -5.65 1.64 -3.44 157 <.001 (comp w/ lim) contrast 2 -6.25 1.65 -3.78 157 <.001 (comp w/ non) contrast 3 -.60 1.62 -.37 157 .356 (lim w/ non) contrast 4 -4.00 1.67 -2.39 157 .009 (comp w/ mod) contrast 5 -2.25 1.65 -1.36 157 .087 (mod w/ non) Hypothesis 3 enjoys full statistical support, while hypothesis 3a enjoys no statistical support. It is significant to note that all means are in the directions predicted. However, not all mean differences are 133 statistically significant. Contrasts 1, 2, and 4 enjoy a high level Of statistical significance indicating that complete comparative advertising has a significantly more negative impact on targeted candidate evaluations than limited comparative advertising, moderate comparative advertising, and noncomparative advertising. Contrasts 3 and 5 are not statistically significant. While the direction and pattern Of change scores indicate support for the linear relationship predicted by hypothesis 3a, the mean differences are not significant. In this instance, complete comparative advertising encourages greater negative change in targeted candidate evaluations than moderate, limited, and noncomparative advertising, as predicted. Moderate and limited comparative advertising, however, are not significantly different from noncomparative advertising in their impact on targeted candidate evaluations. Hypothesesmzandfiomrial Hypothesis 4 proposes that complete comparative advertising will produce greater negative change in evaluations of targeted candidate liking than moderate, limited, and noncomparative advertising. Hypothesis 4a proposes that moderate and limited comparative advertising will produce greater negative change in evaluations of targeted candidate liking that noncomparative advertising (COMP > MOD, LIM > NON). In order to evaluate hypotheses 4 and 4a, pretest candidate liking means were subtracted from 134 posttest candidate liking means to create change scores. The differences between treatment change scores were initially tested using an omnibus F test. The results of analysis of variance suggest that no statistically significant differences exist between treatment means (E (3, 160) = 1.45, p = .23), as indicated in Table 12. Table 12 ANOVA for targeted candidate liking $99293 DE DE MD E P between groups 3 6.56 2.188 1.45 .23 within groups 160 240.56 1.50 total 163 247.12 Despite the lack of overall significance, individual specified contrasts between treatment means are still appropriate, as previously discussed (Keppel, 1973; Ronsethal and Rosnow, 1985; Winer, 1971). For this reason, hypotheses 4 and 4a were tested by specifying orthogonal polynomial contrasts (see Appendix E for weighting specifications). An examination of specified contrasts indicates that no statistically significant differences exist among treatment change scores, as exhibited in Table 13. 135 Table 13 Contrasts for targeted candidate liking a ts Valuo SE 1 DE p contrast 1 —.34 .27 -1.27 160 .10 (comp w/ lim) contrast 2 -.34 .27 -l.27 160 .10 (comp w/ non) contrast 3 .00 .26 .00 160 .95 (lim w/non) contrast 4 .10 .27 .36 160 .64 (comp w/ mod) contrast 5 -.44 .27 -1.64 160 .052 (mod w/ non) Statistical support for hypotheses 4 and 4a does not exist at the .05 level. The results of contrasts l, 2, and 5 indicate that mean differences are in the direction predicted but are not statistically significant. Contrasts 3 and 4 do not exhibit the directionality predicted and so possess an extremely low level of statistical probability. In addition, post hoc analysis appears unlikely to offer any further information concerning significant change score differences addressed by hypotheses 4 and 4a. With this in mind, the predicted effects of comparative advertising on targeted candidate liking are unsubstantiated. Hymthesesflixeandfiixeial Hypothesis 5 proposes that complete comparative advertising will produce greater negative change in targeted candidate voting preferences than moderate, limited, or noncomparative advertising. Hypothesis 5a proposes that 136 moderate and limited comparative advertising will produce greater negative change in targeted candidate voting preferences than noncomparative advertising (COMP > MOD, LIM > NON). In order to address hypotheses 5 and 5a, pretest voting preference means were subtracted from posttest voting preference means to create change scores. The differences between change scores were tested for statistical significance using an omnibus F test. The results of analysis of variance indicate that no statistically significant differences exist between treatment means (F (3, 160) = .93, p = .43), as exhibited in Table 14. Table 14 ANOVA for targeted candidate voting preferences Seuree DE 55 M§ E 9 between groups 3 6.73 2.24 .93 .43 within groups 160 386.39 2.41 total 163 393.12 As previously discussed, it is appropriate to examine planned contrasts between treatment means despite the lack of statistically significant mean differences (Keppel, 1973; Ronsethal and Rosnow, 1985; Winer, 1971). Therefore, orthogonal polynomial contrasts were specified for testing following the application of an omnibus F test (see Appendix E for weighting specifications). An examination of specified contrasts suggests that no significant differences exist among treatment means, as indicated in Table 15. 137 Table 15 Contrasts for targeted candidate voting preferences gourmets Karel 513. 3; DE 2 contrast 1 -.56 .34 -1.64 160 .052 (comp w/ lim) contrast 2 -.23 .34 -.67 160 .25 (comp w/ non) contrast 3 .33 .34 .98 160 .84 (lim w/ non) contrast 4 -.20 .35 -.57 160 .28 (comp w/ mod) contrast 5 -.03 .34 -.08 160 .46 (mod w/ non) There is no statistical support for hypotheses 5 and 5a at the .05 level. Contrasts 1, 2, 4, and 5 contain mean differences in the directions predicted but are not statistically significant. Contrast 3 has a very low level of statistical reliability because the mean is not in the direction predicted. As was the case with targeted candidate liking, the predicted effects Of comparative advertising on targeted candidate voting preferences are unsubstantiated. As an additional note, post hoc analyses conducted on targeted candidate voting preferences do not appear promising given the range Of contrast reliabilities. Post Hoc Analysis of Hypotheses Post hoc analyses are conducted in order to gain a complete understanding of potentially significant treatment condition differences. These analyses are particularly 138 important for contrasts containing score differences in the opposite directions of those predicted. As previously discussed, the Scheffe method is appropriate to use in post hoc analyses because it provides a conservative measure of protection against Type I error for the pairwise comparison of means (Keppel, 1973; Snedecor and Cochran, 1967; Winer, 1971). Hypo—theeeegneandenefel Hypothesis 1 proposes that moderate and limited comparative advertising will be perceived as more fair and unbiased than noncomparative and complete comparative advertising. Hypothesis 1a proposes that noncomparative advertising will be perceived as more fair and unbiased than complete comparative advertising. The fair and unbiased items were combined with believable to form the information credibility subscale. The results of contrast analysis indicate a modest level of support for hypothesis 1 and full support for hypothesis 1a. An omnibus F test was conducted for purposes of post hoc analysis, utilizing all conditions and the control group. The purpose of this analysis was to examine the possibility of statistically significant mean differences in directions Opposite of those predicted. The results of analysis Of variance indicated the existence of significant mean differences (F (4, 201) = 19.27, p <.001), as exhibited in Table 16. 139 Table 16 ANOVA for post hoc analysis Of information credibility Seeree DE DE HE E 2 between groups 4 983.01 245.75 19.27 <.001 within groups 201 2562.54 12.75 total 205 3545.55 A Scheffe’s multiple comparison test was used to determine the location of significant mean differences. A number of significant differences emerged, as indicated in Table 17. Table 17 Mean ratings of information credibility senditien m_ean complete comparative 9.77* limited comparative 11.54* moderate comparative 11.67* noncomparative 14.09** control 15.97 * Significantly different from the control at the .05 level. ** Significantly different from complete and limited comparative advertising at the .05 level. The results of Scheffe’s multiple comparison test, along with the pattern of mean ratings for information credibility, indicate that noncomparative advertising is 140 perceived as being significantly higher in information credibility than both complete comparative advertising and limited comparative advertising. When combined with the analysis of contrasts specified for hypotheses 1 and 1a, these findings indicate that moderate comparative advertising, limited comparative advertising, and noncomparative advertising are all perceived as being significantly higher in information credibility than complete comparative advertising. In addition, noncomparative advertising is perceived as being significantly higher in information credibility that limited comparative advertising. These findings provide no additional support for hypothesis 1 or hypothesis 1a. Wmmm Hypothesis 2 proposes that complete comparative advertising will be perceived as more informative and useful than moderate, limited, and noncomparative advertising. Hypothesis 2a proposes that moderate and limited comparative advertising will be perceived as more informative and useful than noncomparative advertising. The informative and useful items were combined with interesting and believable to form the information utility subscale. The results of contrast analysis indicate little statistical support for hypothesis 2. Although complete comparative advertising is perceived as containing significantly more information utility than limited comparative advertising, none of the other contrasts are significant at the .05 level. In addition, hypothesis 141 2a is not statistically supported. An omnibus F test was conducted for purposes of post hoc analysis. This test utilized all conditions and the control group in order to examine statistically significant mean differences. The results of analysis of variance indicated the existence of significant mean differences (F (4, 201) = 3.09, p = .017), as exhibited in Table 18. Table 18 ANOVA for post hoc analysis of information utility 599292 DE DE HE E D between groups 4 272.41 68.10 3.09 .017 within groups 201 4433.00 22.05 total 205 4705.41 A Scheffe’s multiple comparison test was used in order to determine the location of significant differences. The results of the test indicated that no mean differences existed that were statistically significant at the .05 level of reliability. It is worth noting that a less conservative test of pairwise mean comparisons would likely reveal significant mean differences. Given the extremely conservative nature of Scheffe’s test, however, no mean differences exhibited statistical significance. This post hoc analysis contributes no additional information to the contrasts of means already conducted. Hypothesis 2 possesses little statistical support and hypothesis 2a possesses no statistical support. 142 Hypotheses Three and Three(a) Hypothesis 3 proposes that complete comparative advertising will produce greater negative change in evaluations of the targeted candidate than moderate, limited, and noncomparative advertising. Hypothesis 3a proposes that moderate and limited comparative advertising will produce greater negative change in evaluations Of the target than noncomparative advertising. The results of contrast analysis indicate full support for hypothesis 3 and no support for hypothesis 3a. In this instance, post hoc analysis is unnecessary. The Scheffe’s method will not uncover any additional mean differences that will be statistically significant since this test provides a more conservative level of protection against Type I error than the contrast analysis already used (Keppel, 1973, Snedecor and Cochran, 1967, Winer, 1971). In addition, all contrast differences are in the directions predicted so no differences are statistically significant in directions Opposite of those predicted. For these reasons, post hoc analysis concerning hypotheses 3 and 3a is unnecessary. meheeeeiourendmnel Hypothesis 4 proposes that complete comparative advertising will produce a greater negative change in evaluations of the targeted candidate than moderate, limited, or noncomparative advertising. Hypothesis 4a 143 proposes that moderate and limited comparative advertising will produce a greater negative change in targeted candidate liking than noncomparative advertising. In order to test these hypotheses, pretest evaluations of targeted candidate liking were subtracted from posttest evaluations of targeted candidate liking to create change scores. The results of contrast analysis indicate no statistically significant support of hypotheses 4 or 4a at the .05 level. An omnibus F test was conducted for purposes of post hoc analysis. This analysis is of potential benefit in examining contrast differences in directions opposite of those predicted. The results of analysis of variance were significant (F (4, 201) = 5.18, p <.001), as indicated in Table 19. Table 19 ANOVA for post hoc analysis of targeted candidate liking Seuree DE DE HE E 2 between groups 4 28.04 7.01 5.18 <.001 within groups 202 273.21 1.35 total 206 301.25 A Scheffe’s multiple comparison test was used to determine the location of significant differences. The results Of the test revealed significant change score differences, as indicated in Table 20. 144 Table 20 Change scores for targeted candidate liking Densitien ehememe moderate comparative -1.32* complete comparative -1.22* moderate comparative -.88 noncomparative -.88 control -.28 *Significantly different from the control at the .05 level. Unfortunately, the results of the Scheffe’s test are Of limited use given the purpose of this analysis. While moderate and complete comparative advertising produce significantly greater negative change than the control in evaluations of targeted candidate liking, these findings provide no additional information concerning specific contrast differences. Hypotheses 4 and 4a remain unsupported. mnemzixeandfiiletel Hypothesis 5 predicts that complete comparative advertising will produce greater negative change in targeted candidate voting preferences than moderate, limited, or noncomparative advertising. Hypothesis 5a predicts that moderate and limited comparative advertising will produce greater negative change in targeted candidate voting preferences than noncomparative advertising. In order to test these hypotheses, pretest evaluations Of targeted 145 candidate voting preferences were subtracted from posttest evaluations of targeted candidate voting preferences to create change scores. The results of contrast analysis indicate no statistically significant support of hypotheses 5 or 5a at the .05 level. An omnibus F test was conducted for purposes of post hoc analysis. The results of analysis of variance indicated the existence of significant change score differences (E (4, 202) = 4.16, p = .003), as exhibited in Table 21. Table 21 ANOVA for post hoc analysis of targeted candidate voting preferences DQQIQQ DE DE HE E 2 between groups 4 36.43 9.11 4.16 .003 within groups 202 441.55 2.18 total 206 477.98 A Scheffe’s multiple comparison test was used to determine the exact location Of significant mean differences. The results of the test revealed a significant difference between the complete comparative advertising score and the control group score, as indicated in Table 22. 146 Table 22 Change scores for targeted candidate voting preferences Conditism engagescere complete comparative -1.32* moderate comparative -1.12 noncomparative -1.09 limited comparative -.76 control -.14 *Significantly different from the control at the .05 level. Once again, the results of the Scheffe’s test are Of limited use given the purpose of this analysis. While complete comparative advertising produces significantly greater negative change than the control in evaluations of targeted candidate voting preferences, this finding provides no additional information concerning hypotheses 5 or 5a. Research Questions Research questions are asked in instances in which theoretical support is weak, research findings are nonexistent or contradictory, and no clear rationale exists in support of specific hypotheses. Each research question is addressed here in numerical order. Beseamheuestiengne Research question 1 asks if any of the four types of advertising produce positive change in evaluations Of the targeted candidate. Advertising that results in positive target evaluations, rather than negative target evaluations 147 as intended, is indicative of a backlash effect. In order to examine changes in targeted candidate evaluations, change scores were created by summing candidate evaluation scales and subtracting pretest means from posttest means. An examination of scores indicates that in each condition, evaluations of the targeted candidate change in a negative direction. Since none of the advertising stimuli result in positive evaluations Of the targeted candidate, the answer to research question one is negative, and no additional analysis is necessary. Eeeearehgeeeizienm Research question 2 asks if any of the four types of advertising produce positive or negative changes in sponsoring candidate evaluations. In order to answer this question, sponsoring candidate evaluation scores were summed and the pretest mean was subtracted from the posttest mean to create change scores. Change scores were tested using an omnibus F test in order to determine if significant differences existed among conditions. The results of analysis Of variance suggested that condition change scores were significantly different from each other (F (4, 197) = 3.83, p = .005), as indicated in Table 23. 148 Table 23 ANOVA for sponsoring candidate evaluations SQDZQE DE DE HE E 2 between groups 4 820.07 205.02 3.83 .005 within groups 197 10540.52 53.50 total 201 11360.59 A Scheffe’s multiple comparison test was used to determine the location of significant change score differences. As before, the Scheffe’s test is appropriate because it provides a high level of protection against Type I error for post hoc analysis (Keppel, 1973; Snedecor and Cochran, 1967; Winer, 1971). The results of the Scheffe’s test revealed the existence of significant differences, as indicated in Table 24. Table 24 Change scores for sponsoring candidate evaluations gonditien ehmgeeeere limited comparative -3.95 complete comparative -2.15 control -1.49 noncomparative 1.21* moderate comparative 1.29* * Significantly different from limited comparative advertising at the .05 level. The results Of the Scheffe’s test indicate that 149 sponsoring candidate evaluations are significantly more positive in the noncomparative and moderate comparative advertising conditions than in the limited comparative advertising condition. NO other change score differences are significant at the .05 level. When examining change score differences, it is interesting to note the relatively large, negative change in control group evaluations of the sponsoring candidate. This negative change is identified by Campbell and Stanley (1963) as testing; the effect of taking a test on subsequent test scores. Despite the random assignment of subjects to conditions which helps to correct for testing imbalances (Campbell and Stanley, 1963), this appears to be a concern in the instance of sponsoring candidate evaluations. In order to determine if the negative change in sponsoring candidate evaluations was significant, the pretest sponsoring candidate mean for the control group was compared to the posttest sponsoring candidate mean for the control group using a one-sample t-test. The results of the test indicated that the difference between the pretest measure and posttest measure for the control group was not statistically significant at the .05 level (t = -1.23, p = .23). Despite this finding, additional tests were undertaken as a precaution, given the relatively large testing effect. In this instance, treatment condition change scores were tested using one-sample t-tests. By testing against the change score of the control group, the effects of testing 150 are statistically reduced to zero and testing effects are controlled. The results Of this analysis indicated that both the moderate comparative advertising score (t = 2.08, p = .044) and the noncomparative advertising score (t = 3.17, p = .003) were significant when the effects of testing were controlled. These findings provide an additional level of protection from making a Type I error concerning the significance of differences in change scores between treatment conditions. In answer to research question 2 then, it appears as if both moderate comparative advertising and noncomparative advertising significantly increase sponsoring candidate evaluations, when compared to the decrease in sponsoring candidate evaluations produced by limited comparative advertising. Additional change score differences were not significant at the .05 level. EesearehQuestienTbree Research question 3 asks if any of the four types of advertising produce positive changes in sponsoring candidate liking. In order to answer this question, change scores were created by subtracting pretest ratings of sponsoring candidate liking from posttest ratings of sponsoring candidate liking. An examination of means indicated that only noncomparative advertising contributed to positive sponsoring candidate evaluations. An omnibus F test was used to determine if the positive noncomparative score was significantly different from the remaining negative scores. 151 The results Of analysis of variance indicated that no significant differences existed between treatment condition scores (E (4, 201) = 1.49, p = .20). An examination of change scores, however, revealed that the control score fell .50 as a result of testing. As a precaution, a one-sample t-test was conducted between the control group’s pretest rating of sponsor liking and posttest rating of sponsor liking. Despite its relatively small size, the results of the t-test indicated that the change was statistically significant (t = -1.69, p = .005). The results of a one-sample t-test using the control group change score to remove the effects of testing, indicated that the positive change in the noncomparative advertising condition was significant (t =5.00, p = <.001). This finding indicates that when the effects of testing are controlled, noncomparative advertising significantly increases perceptions of sponsoring candidate liking. As a final note, the remaining treatment means did not require additional testing since none of the scores produced positive changes in evaluations of sponsoring candidate liking. Beeserehguestienmur. Research question 4 asks if any of the four types of advertising produce positive changes in evaluations Of sponsoring candidate mean spiritedness. In this instance, greater amounts of mean spiritedness are represented by higher subject ratings and would be indicative Of a backlash L—— 152 effect. With this in mind, pretest ratings of sponsoring candidate mean spiritedness were subtracted from posttest ratings of sponsoring candidate mean spiritedness in order to create change scores. All change scores showed increases in ratings of mean spiritedness with the exceptions of noncomparative advertising which decreased slightly (-.40) and moderate comparative advertising which did not change. An omnibus F test was used in order to determine if significant differences existed between scores. The results Of analysis Of variance indicated that scores were significantly different from each other (E (4, 201) = 5.42, p <.001), as indicated in Table 25. Table 25 ANOVA for sponsoring candidate mean spiritedness surge 21: _se a z :2 between groups 4 42.86 10.71 5.42 <.001 within groups 201 397.16 1.97 total 205 440.02 A Scheffe’s multiple comparison test was used in order to determine the location of significant change score differences. The results of the Scheffe’s test indicated the existence of significant differences, as exhibited in Table 26. 153 Table 26 Change scores for sponsoring candidate mean spiritedness gondition c an ooopo noncomparative -.40 limited comparative .00 control .21 moderate comparative .27 complete comparative 1.00* * Significantly different from noncomparative advertising and limited comparative advertising at the .05 level. The Scheffe’s test indicates that complete comparative advertising significantly increases perceptions of sponsoring candidate mean spiritedness when compared with perceptions of candidate mean spiritedness in the noncomparative and limited comparative advertising conditions. As with other analyses, a testing effect is evident in the change score of the control group. As a precaution, the pretest mean of the control group was compared with the posttest mean of the control group using a one-sample t- test. The results of the test indicated that the difference was not statistically significant (t = .86, p = .39). No additional analyses were conducted given the lack Of statistical significance and relatively small size of the testing effect. 154 Beseareh Question Eire Research question 5 asks if any of the four types Of advertising produce positive or negative changes in voting preferences for the sponsoring candidate. In order to examine changes in sponsoring candidate voting preferences, change scores were created by subtracting pretest sponsoring candidate voting means from posttest sponsoring candidate voting means. An examination of change scores indicates that advertising stimuli produced both increases and decreases in voting preferences for the sponsoring candidate. An omnibus F test was used in order to determine if the differences among group means were statistically significant. The results of analysis of variance indicated that no change scores were significantly different at the .05 level (E (4, 200) = 1.99, p = .09). Once again, the control group change score evidenced a small testing effect (-.16). As a precaution, the pretest mean of the control group was compared with the posttest mean of the control group using a one-sample t-test. The results of the test indicated that the difference was not statistically significant (t = -.69, p = .49). No additional analyses were conducted given the lack of statistical significance and relatively small size Of the testing effect. 155 Web Limest' §_xi Research question six asks if moderate and limited comparative advertising differ in their abilities to: Produce positive change in sponsoring candidate evaluations or negative change in targeted candidate evaluations; produce positive change in sponsoring candidate liking or negative change in targeted candidate liking; produce increases in perceptions of sponsoring candidate mean spiritedness; produce positive change in sponsoring candidate voting preferences or negative change in targeted candidate voting preferences; and, produce positive change in election involvement. Moderate and limited comparative advertising were tested for different effects on dependent variables using independent samples t-tests. In order to determine if moderate and limited comparative advertising differed in their abilities to produce positive change in sponsoring candidate evaluations, or negative change in targeted candidate evaluations, change scores were created by summing candidate evaluation scales and subtracting pretest means from posttest means. The results of independent samples t- tests indicated that moderate comparative advertising produced a significant positive change in aggregate evaluations of the sponsoring candidate when compared with the negative change produced by limited comparative advertising (t = 2.80, p = .007). In the instance of targeted candidate evaluations, no statistically significant 156 difference existed between moderate and limited comparative advertising (t = -.98, p = .32). Moderate and limited comparative advertising were tested to determine if they differed in their abilities to produce positive change in sponsoring candidate liking, or negative change in targeted candidate liking. Pretest measures of candidate liking were subtracted from posttest measures of candidate liking to create change scores. In the instance sponsoring candidate liking, the results of an independent samples t-test indicated that the change score differences of moderate and limited comparative advertising were not significantly different from each other (t = 1.12, p = .27). In the same way, a t-test indicated that moderate and limited comparative advertising did not significantly differ in their impact on targeted candidate liking (t = -l.56, p = .12). Moderate and limited comparative advertising were tested to determine if they differed in their abilities to produce increases in perceptions Of sponsoring candidate mean spiritedness, following tests of candidate liking. Pretest measures of sponsor mean spiritedness were subtracted from posttest measures Of sponsor mean spiritedness to create change scores. An independent samples t-test was used to determine if the change score differences for moderate and limited comparative advertising were significantly different from each other. The results of the test indicated that the change score differences between moderate and limited comparative advertising were 157 not significantly different (t = .92, p = .36). Moderate and limited comparative advertising were tested to determine if they differed in their abilities to produce positive change in voting preferences for the sponsoring candidate, or negative change in voting preferences for the targeted candidate. Pretest measures of candidate voting preferences were subtracted from posttest measures of candidate voting preferences to create change scores. The results of an independent samples t-test indicated that moderate and limited comparative advertising were not significantly different in their abilities to produce positive change in sponsor voting preferences (t = 1.48, p = .14). In the same way, a t-test indicated that moderate and limited comparative advertising did not significantly differ in their abilities to produce negative change in targeted candidate voting preferences (t = -1.04, p = .30). Finally, moderate and limited comparative advertising were tested to determine if they differed in their abilities to produce positive change in election involvement. Pretest measures of election involvement were subtracted from posttest measures of election involvement to create change scores. An independent samples t-test was used to determine if the change score differences for moderate and limited comparative advertising were significantly different from each other. The results of the test indicated that moderate and limited comparative advertising did not significantly 158 differ in their abilities to produce positive change in election involvement (t = .79, p = .43). In brief review of research question 6, the results of independent samples t-tests indicate that moderate comparative advertising produces a significant positive change in aggregate evaluations of the sponsoring candidate when compared with the negative change produced by limited comparative advertising. In all other instances, however, there are no significant differences between moderate and limited comparative advertising. ResearshQuestisOSesLen Research question 7 asks if any of the four types of advertising produce positive changes in election involvement. In order to address this question, election involvement scales were summed and pretest means were subtracted from posttest means to create change scores. Change scores were tested using an omnibus F test in order to determine if significant differences existed. The results Of analysis Of variance suggested that change scores were significantly different from each other (E (4, 202) = 6.98, p < .001), as indicated in Table 27. 159 Table 27 ANOVA for election involvement EDDIQE DE DE HE E 2 between groups 4 512.30 128.07 6.98 <.001 within groups 202 3703.44 18.33 total 206 4215.74 A Scheffe’s multiple comparison test revealed the location of significant differences, as indicated in Table 28. Table 28 Change scores for election involvement sensation shengessere noncomparative -1.38 control —1.37 limited comparative .07 moderate comparative .90 complete comparative 2.85* * Significantly different from noncomparative adverting and the control group at the .05 level. The results of the Scheffe’s test indicate that election involvement is significantly increased by complete comparative advertising when compared with the decreases in involvement existing in the noncomparative advertising condition and the control group. No other change score differences are significant at the .05 level. 160 When examining change score differences, the relatively large, negative change in control evaluations of election involvement should be noted. As before, this negative change is due to the effects of testing. In order to determine if the negative change in sponsoring candidate evaluations was significant, the pretest election involvement score for the control group was compared to the posttest election involvement score for the control group using a one-sample t-test. The results of the test indicated that the difference between the pretest measure and posttest measure for the control group was not statistically significant at the .05 level (t = -1.69, p = .09). Despite this finding, additional tests were undertaken as a precaution, given the relatively large testing effect. As before, treatment condition change scores were tested for significance using the control group change score to control for the effects of testing. The results of one—sample t- tests indicated that changes in both the complete comparative advertising score (t = 5.81, p <.001) and the moderate comparative advertising score (t = 3.59, p = .001) were significant. Additional change score differences were not significant at the .05 level. Summary of Findings Research findings are summarized here for ease of reference. Findings are grouped by dependent variable and include the results of hypothesis testing, post hoc analysis and the analysis of research questions. 161 Infermtien £3dele Hypothesis 1 proposes that moderate and limited comparative advertising will be perceived as more fair and unbiased than noncomparative and complete comparative advertising. Hypothesis 1a proposes that noncomparative advertising will be perceived as more fair and unbiased than complete comparative advertising. Fair and unbiased were combined with believable to form an information credibility subscale. The results of contrast analysis indicate that noncomparative advertising, limited comparative advertising, and moderate comparative advertising possess significantly more information credibility than complete comparative advertising. Moderate and limited comparative advertising do not possess significantly more information credibility than noncomparative advertising, however. These findings indicate moderate support for hypothesis 1 and full support for hypothesis 1a. Mean differences concerning information credibility were submitted to post hoc analysis in order to examine the possibility of significant mean differences in directions opposite of those predicted. The results of this analysis reveal that noncomparative advertising possesses significantly more information credibility than both complete comparative advertising and limited comparative advertising. The combined results of both analyses, along with the pattern of mean ratings, indicate that 162 noncomparative advertising, moderate comparative advertising, and limited comparative advertising possess significantly more information credibility than complete comparative advertising. In addition, noncomparative advertising possesses significantly more information credibility than limited comparative advertising. I 50 !' HI'l't Hypothesis 2 proposes that complete comparative advertising will be perceived as more informative and useful than moderate, limited, and noncomparative advertising. Hypothesis 2a proposes that moderate and limited comparative advertising will be perceived as more informative and useful that noncomparative advertising. Informative and useful were combined with interesting and believable to form an information utility subscale. The results of contrast analysis indicate that complete comparative advertising possesses significantly more information utility than limited comparative advertising. The remaining contrasts are not statistically significant. Statistical support for hypothesis 2 is weak and there is no statistical support for hypothesis 2a. Mean differences concerning information utility were submitted to post hoc analysis in order to examine the possibility of significant mean differences in directions opposite of those predicted. The results Of this analysis indicate that no mean differences are statistically significant at the .05 level of probability. These findings 163 contribute no additional information concerning hypotheses 2 and 2a. Aggregate gandioate Evalnotions Hypothesis 3 predicts that complete comparative advertising will produce greater negative change in evaluations of the targeted candidate than moderate, limited, and noncomparative advertising. Hypothesis 3a predicts that moderate and limited comparative advertising will produce greater negative change in evaluations Of the targeted candidate than noncomparative advertising. Candidate evaluations were summed and the pretest mean was subtracted from the posttest mean to create change scores. The results of contrast analysis indicate that complete comparative advertising has a significantly more negative impact on targeted candidate evaluations than moderate comparative advertising, limited comparative advertising, and noncomparative advertising. While all other change scores are in the directions predicted, no other change score differences are significant. Post hoc analysis is unnecessary. These findings indicate full support for hypothesis 3 and no support for hypothesis 3a. Research question 1 asks if any of the four types of advertising produce positive changes in targeted candidate evaluations. An examination of scores indicates that all evaluations of the targeted candidate change in a negative direction. Therefore, the answer to research question 1 is negative. 164 Research question 2 asks if any of the four types of advertising produce positive or negative changes in sponsoring candidate evaluations. An analysis of change scores indicates that significant differences exist between treatment conditions and the control group. The results of a Scheffe’s multiple comparison test indicate that sponsoring candidate evaluations are significantly more positive in the noncomparative advertising condition and moderate comparative advertising condition than in the limited comparative advertising condition. Treatment condition change scores were tested using the control group change score as a means of controlling for the effects of testing. The results of one-sample t-tests indicate that both moderate comparative advertising and noncomparative advertising significantly increase sponsoring candidate evaluations. ; 1.: ! I'l' Hypothesis 4 suggests that complete comparative advertising will produce greater negative change in targeted candidate liking than moderate, limited, and noncomparative advertising. Hypothesis 4a suggests that moderate and limited comparative advertising will produce greater negative change in targeted candidate liking than noncomparative advertising. Pretest means of targeted candidate liking were subtracted from posttest means of targeted candidate liking to create change scores. The results of contrast analysis indicate a lack of statistical 165 support for hypotheses 4 and 4a at the .05 level. Post hoc analyses were conducted to examine potentially significant contrasts in directions opposite of those predicted. The results Of analysis of variance indicate that significant change score differences exist. The results Of a Scheffe’s test indicate that both the moderate and complete comparative advertising change scores are significantly different from the change score of the control group at the .05 level. These findings provide no additional information concerning hypotheses 4 and 4a. Research question 3 asks if any of the four types of advertising produce positive changes in sponsoring candidate liking. Only noncomparative advertising produces a positive change in sponsoring candidate liking. While an analysis of variance indicates that no significant differences exist between treatment condition change scores, additional analyses reveal a significant testing effect. The noncomparative advertising change score was examined while controlling for the effects of testing. The result of this analysis indicates that noncomparative advertising significantly increases sponsoring candidate liking. Research question 4 asks if any of the four types of advertising produce a positive change in perceptions of sponsoring candidate mean spiritedness. In this instance, a positive change indicates that perceptions of candidate mean 166 spiritedness increase among subjects. Pretest measures of candidate mean spiritedness were subtracted from posttest measures of candidate mean spiritedness to create change scores. Analysis of variance indicates that change scores are significantly different from each other. The results of a Scheffe’s multiple comparison test reveals that complete comparative advertising significantly increases perceptions of sponsoring candidate mean spiritedness when compared with perceptions in the noncomparative and limited comparative advertising conditions. NO additional analyses were conducted. Qandidefeyotingzreferense Hypothesis 5 predicts that complete comparative advertising will produce a greater negative change in targeted candidate voting preferences than moderate, limited, or noncomparative advertising. Hypothesis 5a predicts that moderate and limited comparative advertising will produce a greater negative change in targeted candidate voting preferences than noncomparative advertising. Pretest voting preference means were subtracted from posttest voting preference means to create change scores. The results Of contrast analysis indicate that there is no statistical support of hypotheses 5 or 5a at the .05 level. Post hoc analyses were conducted to examine the possibility of significant contrasts in directions opposite of those predicted. The results of analysis of variance 167 indicate that significant change score differences exist. The results of a Scheffe’s test reveals a significant difference between the complete comparative advertising condition score and the control group score. This finding provides no additional information concerning hypotheses 5 or 5a. Research question 5 asks if any of the four types of advertising produce positive or negative changes in voting preferences for the sponsoring candidate. An examination Of change scores indicates that advertising stimuli produce both increases and decreases in voting preferences for the sponsoring candidate. The results of analysis of variance, however, indicate that no change scores are significantly different at the .05 level. Research question 6 asks if moderate and limited comparative advertising differ in their abilities to: Produce positive change in sponsoring candidate evaluations or negative change in targeted candidate evaluations; produce positive change in sponsoring candidate liking or negative change in targeted candidate liking; produce increases in sponsoring candidate mean spiritedness; produce positive change in sponsoring candidate voting preferences or negative change in targeted candidate voting preferences; and, produce positive change in election involvement. In order to address these questions, moderate and limited comparative advertising were tested for significant 168 differences using independent samples t-tests. The results of these tests indicate that moderate comparative advertising produces a significant positive change in evaluations of the sponsoring candidate when compared with the negative change produced by limited comparative advertising. In all other instances, however, there are no significant differences between moderate and limited comparative advertising. Eleetionlnxelxement Research question 7 asks if any of the four types of advertising produce positive changes in election involvement. Measures of election involvement were summed and the pretest mean was subtracted from the posttest mean to create change scores. Analysis Of variance indicates the existence of statistically significant change score differences. A Scheffe’s multiple comparison test was used to determine the location Of significant change score differences. The results of the Scheffe’s test indicate that complete comparative advertising significantly increases election involvement when compared with the decreases in involvement in the noncomparative advertising condition and the control group. NO other change score differences are significant at the .05 level. In order to control for the possible effects of testing, treatment condition change scores were tested against the control group change score using one-sample t- tests. The results Of this analysis indicate that both 169 complete comparative advertising and moderate comparative advertising are significantly different from the control group. Additional change score differences are not significant at the .05 level. CHAPTER FIVE DISCUSSION AND CONCLUSIONS The research findings discussed here include the results of hypothesis testing and post hoc analysis, followed by the analysis of research questions. The limitations of the study are addressed next, followed by recommendations for a course Of future research and a summary Of conclusions. Discussion and Implications of Research Findings The results of the experiment and their implications are discussed here. All results are presented briefly in summary form, followed by a brief a discussion of the findings and their implications. mneses Hypothesis 1 proposed that moderate and limited comparative advertising are perceived as more fair and unbiased than noncomparative and complete comparative advertising. Hypothesis 1a proposed that noncomparative advertising is perceived as more fair and unbiased than complete comparative advertising. Fair and unbiased were combined with believable to form an information credibility subscale. The results of contrast analysis indicated that noncomparative advertising, limited comparative advertising, 170 171 and moderate comparative advertising possessed significantly more information credibility than complete comparative advertising. No other mean differences were significant. These results indicate partial support for hypothesis 1 and full support for hypothesis 1a. Post hoc analyses were conducted. The results of a Scheffe’s test revealed that noncomparative advertising possessed significantly more information credibility than both complete comparative advertising and limited comparative advertising. Taken together, these findings indicate that moderate, limited, and noncomparative advertising possess significantly more information credibility than complete comparative advertising. In addition, noncomparative advertising possesses significantly more information credibility than complete comparative advertising. These results are generally consistent with the results of a variety of survey research indicating that voters dislike negative political advertising, finding it uninformative, unethical, uninteresting, and deceptive (Garramone, 1984; Johnson-Cartee and Copeland, 1989: Nix, Nein, 1984; Pinkleton and Garramone, 1992; Surlin and Gordon, 1977). These findings are of particular interest in view of the increased use of comparative political advertising by candidates attempting to avoid the negative stigma attached to attack advertising, as previously discussed. In this instance, the negativity Of complete 172 comparative advertising significantly lowered its information credibility, despite its comparative format. Moderate comparative advertising did not suffer the same loses of information credibility, however, despite the fact that it is only slightly less negative. These findings suggest that voters are likely to recognize and distrust the most negative kinds of political advertising, regardless of their form. At the same time, political advertising can be negative and still enjoy an acceptable degree Of information credibility. With this in mind, candidates for public office may find it beneficial to employ comparative appeals containing moderate amounts of negative information. This strategy will allow them to communicate disparaging information regarding an opponent to voters while maintaining an acceptable level of information credibility. Hypothesis 2 proposed that complete comparative advertising is perceived as more informative and useful than moderate, limited, and noncomparative advertising. Hypothesis 2a proposed that moderate and limited comparative advertising are perceived as more informative and useful that noncomparative advertising. Informative and useful were combined with interesting and believable to form an information utility subscale. The results of contrast analysis indicated that complete comparative advertising possessed significantly more information utility than limited comparative advertising. The remaining contrasts were not statistically 173 significant. Post hoc analyses were conducted. The results Of a Scheffe’s test indicated that no mean differences were statistically significant at the .05 level of probability. Statistical support for hypothesis 2 is weak and there is no statistical support for hypothesis 2a. Despite the general lack of support for hypotheses 2 and 2a, the significant difference found between complete and limited comparative advertising is broadly consistent with other research findings. The results of the Johnson- Cartee and Copeland (1991a) study, as well as the results of a limited amount of comparative product advertising research utilizing print media, indicate that comparative advertising is generally more informative than other types Of advertising. In the present study, complete comparative advertising contained the greatest amount of comparative information, while the limited stimulus contained very little comparative information. Here, the stimulus with the most comparative information enjoyed significantly more information utility than a stimulus possessing much less comparative information. It should be noted, however, that the information utility of complete comparative advertising appears to be weak. It is likely that the limited comparative stimulus communicated very little new information concerning candidate differences to subjects. This is due to the presence of only two comparative statements differentiating the sponsoring candidate from the targeted candidate in the limited advertisement. Given this fact, the information 174 utility of complete comparative advertising is possibly a function of the lack of information utility possessed by limited comparative advertising rather than the true information utility possessed by complete comparative advertising. Additional support for this position is found in the fact that no additional significant differences emerged between the complete comparative stimulus and other stimuli. Hypothesis 3 predicted that complete comparative advertising produces a greater negative change in evaluations Of the targeted candidate than moderate, limited, and noncomparative advertising. Hypothesis 3a predicted that moderate and limited comparative advertising produce a greater negative change in evaluations of the targeted candidate than noncomparative advertising. The results Of contrast analysis indicated that complete comparative advertising had a significantly more negative impact on targeted candidate evaluations than moderate comparative advertising, limited comparative advertising, and noncomparative advertising. No other contrasts were significant at the .05 level. These findings indicate complete support for hypothesis 3 and no support for hypothesis 3a. Additional post hoc analyses were unnecessary. Findings in support of hypothesis 3 are completely in keeping with the results of research concerning negative information effects and negative political advertising 175 effects. As previously discussed, research concerning the impact of negative information on person perception indicates that negative information is weighted more heavily than positive information when forming evaluations of social stimuli (Anderson, 1965; Hodges, 1974; Kellerman, 1984, 1989; Levin and Schmidt, 1969; Warr and Jackson, 1976; Wyer, 1970). The results of research concerning the impact of political advertising also indicate that subject evaluations Of targeted candidates Often fall as a result of exposure to negative political advertising (Kaid and Boydston, 1987; Shapiro and Rieger, 1989; Johnson-Cartee and Copeland, 1991a). The fact that complete comparative advertising significantly lowers targeted candidate evaluations indicates that the impact of negative comparative advertising is similar to the impact of other types Of negative political advertising. This finding is particularly interesting giVen the concerns raised regarding the information credibility Of complete comparative advertising. Based on these results, it appears as if complete comparative advertising effectively reduces support of the targeted candidate, despite the fact that it is significantly lower in information credibility than other types of comparative and noncomparative advertising. Although not directly supported by this research evidence, it is reasonable to expect that there is an Optimal tradeoff between the credibility and negativity of political advertising. In this instance, negative 176 advertising will enjoy an acceptable degree of credibility and still effectively reduce support of a targeted candidate. Finally, it is significant to note that all stimuli produced negative changes in targeted candidate evaluations. These findings provide no indication of a backlash effect and will be discussed in greater detail shortly. Hypothesis 4 suggested that complete comparative advertising produces a greater negative change in targeted candidate liking than moderate, limited, and noncomparative advertising. Hypothesis 4a suggested that moderate and limited comparative advertising produce a greater negative change in targeted candidate liking than noncomparative advertising. The results of contrast analysis indicated a lack of statistical support for hypotheses 4 and 4a at the .05 level. Post hoc analyses were conducted. The results of a Scheffe’s test indicated that moderate and complete comparative advertising change scores were significantly different from the control group at the .05 level. These findings provide no additional support for hypotheses 4 or 4a. Based on these results, the impact of negative information on aggregate candidate evaluations and the impact Of negative information on evaluations of candidate liking do not appear to be equivalent. Negative comparative information has a pronounced impact on overall targeted candidate evaluations. At the same time, this information 177 has very little impact on targeted candidate liking. Although improbable, it is possible that negative political advertising impacts voters’ perceptions of a candidate’s character traits and job performance abilities, apart from impacting voters’ evaluations of candidate liking. In this instance, the influence Of complete comparative advertising would be very narrow, significantly reducing targeted candidate evaluations while leaving evaluations of targeted candidate liking intact. It appears more likely that the measure Of candidate liking is not accurately reflective Of subjects’ candidate perceptions. The artificiality of the research environment, the relatively low election involvement of subjects, and the lower reliability of a single-item measure contribute to a situation in which negative information erroneously appears to have no significant impact on targeted candidate liking. Support for this position is found in the general lack of significance existing in contrasts concerning targeted candidate liking. A future study conducted during an election, using a random sample of voters, and utilizing more thorough measures of targeted candidate liking is more likely to accurately measure the impact Of negative information on targeted candidate liking. Hypothesis 5 predicted that complete comparative advertising produces a greater negative change in targeted candidate voting preferences than moderate, limited, or noncomparative advertising. Hypothesis 5a predicted that moderate and limited comparative advertising produce a 178 greater negative change in targeted candidate voting preferences than noncomparative advertising. The results of contrast analysis indicated that there was no statistical support Of hypotheses 5 or 5a at the .05 level of significance. Post hoc analyses were conducted. The results of a Scheffe’s test revealed a significant difference between the complete comparative advertising condition and the control group. This finding provides no additional information concerning hypotheses 5 or 5a, which remain unsupported. As with candidate liking, it appears as if measures of candidate voting preferences are not accurately reflective of subjects’ voting behavior. In this instance, the artificiality of the research environment and the relatively low election involvement of subjects contributes to a situation in which negative information erroneously appears to have no significant impact on subject voting preferences. Support for this position is found in the general lack of significance existing in contrasts concerning targeted candidate voting preferences. As before, a future study conducted during an election and using a random sample of voters is more likely to accurately measure the impact of negative information on targeted candidate voting preferences. Beseershegestiens Research question I asked if any of the four types of advertising produce positive changes in targeted candidate 179 evaluations. An examination of scores indicates that all evaluations of the targeted candidate changed in a negative direction. Therefore, the answer to research question one is negative. This question was included in this study as part of an attempt to gauge the existence Of a backlash effect. The answer to research question 1 indicates that no backlash effect exists as measured by targeted candidate evaluations. It should be noted, however, that targeted candidate evaluations are only a single measure of voter backlash. Candidate liking, candidate voting preferences, and perceived candidate mean spiritedness are also used to provide evidence of a backlash effect. Each of these will be discussed shortly. Research question 2 asked if any of the four types Of advertising produce positive or negative changes in sponsoring candidate evaluations. An analysis of change scores indicated that sponsoring candidate evaluations were significantly more positive in the noncomparative and moderate comparative advertising conditions than in the limited comparative advertising condition. When controlling for the effects of testing, both moderate comparative advertising and noncomparative advertising significantly increased sponsoring candidate evaluations. These results are somewhat difficult to interpret in that both a negative advertisement and a positive advertisement significantly increase sponsoring candidate evaluations. Findings concerning moderate comparative 180 advertising indicate the possibility of a range of acceptability for the contents of comparative political advertising. When comparative advertisements are too negative--as in the instance Of complete comparative advertising-~voters may perceive the advertising as unfair and sponsoring candidate evaluations will decline (see Johnson-Cartee and Copeland, 1989; Garramone, 1984, 1985; Merritt, 1984). In the same way, comparative advertising containing a small amount Of information differentiating between candidates--as in the instance of limited comparative advertising--may frustrate voters by providing little information of use in determining the relative advantages of the sponsoring candidate over the targeted candidate. This also results in declining sponsoring candidate evaluations. An acceptable range of negative and comparative information appears to have been contained in the moderate comparative stimulus. In this instance, enough negative information is presented to effectively erode targeted candidate support. At the same time, the moderate stimulus is not negative enough to create a backlash against the sponsoring candidate. The result is a significant increase in sponsoring candidate evaluations. Findings concerning noncomparative advertising are easier to interpret. In this instance, it appears as if noncomparative advertising is simply having its desired impact. That is, noncomparative advertising is designed to increase positive perceptions of the sponsor, rather than 181 increase negative perceptions of the target. Noncomparative advertising generates no perceptions of unfairness because it communicates only positive information concerning its sponsor. Here, there are no backlash concerns. As an additional note, a backlash effect is weakly evident in sponsoring candidate evaluations. In this instance, noncomparative advertising significantly increases sponsoring candidate evaluations when compared with the decrease in evaluations resulting from limited comparative advertising. Unfortunately, this finding is difficult to interpret because of two related factors. First, it is the least negative of the comparative advertising stimuli which generates the greatest decrease in sponsoring candidate evaluations. Second, another negative stimulus actually has an effect Opposite Of the backlash created by the limited comparative stimulus. These confounding variables contribute a high degree of doubt as to the validity Of the backlash effect created by limited comparative advertising. As previously noted, the primary explanation for this finding exists in the fact that the limited stimulus communicates a small degree of information useful to subjects in determining the advantages of the sponsoring candidate over the targeted candidate. As before, it is likely that subject frustration contributed to a significant decline in sponsoring candidate evaluations. Research question 3 asked if any of the four types of advertising produce a positive change in sponsoring 182 candidate liking. When pretest candidate liking scores were subtracted from posttest candidate liking scores, only one positive score emerged. In addition, the results of analysis of variance indicated that no significant differences existed between treatment condition scores. When controlling for the effects Of testing, however, a one-sample t-test indicated that noncomparative advertising did, in fact, significantly increase sponsoring candidate liking. As with sponsoring candidate evaluations, this finding is probably best interpreted as an example of noncomparative advertising having its desired impact. Since the noncomparative stimulus communicates only positive information about the sponsor Of the advertisement, it is reasonable to expect that sponsor liking will increase, particularly in the absence of any other candidate information. This finding, when combined with findings concerning the positive impact of noncomparative advertising on sponsoring candidate evaluations, indicates that the impact of positive political advertising can be fairly substantial. While negative political advertising appears to generate the majority of attention from journalists and researchers, positive political advertising appears to possess a relatively high degree of effectiveness in achieving its Objectives. Research question 4 asked if any of the four types of advertising produce a positive change in perceptions of 183 sponsoring candidate mean spiritedness. The results of analysis indicated that complete comparative advertising significantly increased perceptions of sponsoring candidate mean spiritedness when compared with noncomparative and limited comparative advertising. These findings provide some evidence of a backlash effect. In this instance, the most negative comparative stimulus contributes to a significant increase in perceptions of sponsoring candidate mean spiritedness. Several points should be made regarding this finding. First, candidate mean spiritedness is only one aspect of the concept of voter backlash. There are a variety of additional measures of voter backlash, as previously discussed. These include sponsoring and targeted candidate evaluations, sponsoring and targeted candidate liking, and sponsoring and targeted candidate voting preferences. Next, it is conceivable for a candidate to engender increasing perceptions of mean spiritedness without having large numbers of the electorate actually vote against him or her. At lower levels, some degree of mean spiritedness is likely to be acceptable for a political candidate. In fact, a certain amount of mean spiritedness may even be beneficial for a candidate attempting to demonstrate a strong image, for example. As voter perceptions of mean spiritedness continue to escalate, however, its consequences are likely to become more significant. Based on the results of this research, complete comparative advertising may harm a 184 candidate by contributing to voter backlash as perceptions Of mean spiritedness reach higher levels. Finally, only the most negative comparative stimulus increased perceptions Of candidate mean spiritedness. Here, it appears as if a less negative form of comparative advertising can be used without creating this perception. As a final note, other indications of voter backlash examined in this study provide minimal evidence of backlash effects. Based on these findings, it appears as if comparative political advertising can be effectively used to communicate negative information to voters concerning an Opponent without engendering the strong backlash effects associated with direct attack advertising. Research question 5 asked if any Of the four types Of advertising produce a positive or negative change in voting preferences for the sponsoring candidate. An analysis of change scores indicated that advertising stimuli produced both increases and decreases in voting preferences for the sponsoring candidate. No change scores were significantly different at the .05 level, however. This result is difficult to interpret, particularly given the lack of consistent research findings concerning the impact of political advertising on voting behavior. As previously discussed, the positions of political consultants concerning political advertising suggest that negative advertising exerts a powerful influence over the attitudes and behavior of voters, generally resulting in its desired effect (Johnson-Cartee and Copeland, 1991a, 1991b; Perloff 185 and Kinsey, 1990). Conversely, the research efforts of academicians examining negative political advertising indicate that a backlash effect is a common outcome Of negative political advertising (Garramone, 1984, 1985; Merritt, 1984; Shapiro and Rieger, 1989). Two points should be made with respect to this finding. First, political scientists have suggested that comparative advertising is a desirable vehicle Of political communication for the very reason that it avoids the stigma attached to negative, attack advertising (Salmore and Salmore, 1989). In providing limited evidence Of a backlash effect, the findings of this study are generally consistent with this view. Second, measures of candidate voting preferences tend to lack a high degree of external validity. Having subjects project their voting behavior in a fictional political contest is quite different from voting for a political candidate in an election. Simply having subjects rate their voting preferences for two candidates on seven-point scales limits the applicability of voting preference measures to actual voting behavior in which one candidate is selected over another. While findings concerning the impact of comparative political advertising on sponsoring candidate voting behavior are not significant in the present study, this should not necessarily be taken as evidence Of null effects. Research question 6 addresses the differences between 186 moderate and limited comparative advertising in terms of their impact on dependent variables. The results of analysis indicated that moderate comparative advertising produced a significant, positive change in aggregate evaluations of the sponsoring candidate when compared with the negative change produce by limited comparative advertising. In all other instances, no significant differences existed between moderate and limited comparative advertising. This finding is interesting in that two stimuli, having no significant difference in terms of perceived negativity, have significant Opposite effects on candidate evaluations. Once again, it should be noted that the limited stimulus contained little comparative information communicating a clear sponsoring candidate advantage. In this instance, the sponsoring candidate may appear to have possessed no real advantage over the targeted candidate resulting in decreased candidate evaluations. In the instance of the moderate stimulus, however, sufficient information exists to allow subjects to identify genuine differences between the sponsoring and targeted candidates. In addition, the moderate stimulus is not so negative as to create a backlash effect. Here, the moderate stimulus successfully increases evaluations of the sponsoring candidate. Further, the sponsoring candidate is exemplified without being excessively negative toward the targeted candidate. Research question 7 asked if any Of the four types Of advertising produce positive changes in election 187 involvement. An analysis of change score differences indicated that complete comparative advertising significantly increased election involvement when compared with the decreases in involvement in the noncomparative advertising condition and the control group. No other change score differences were significant at the .05 level. When controlling for the effects of testing, both the complete and moderate comparative advertising change scores significantly increased election involvement. These findings are particularly interesting in that no previous research exists indicating that political advertising has the ability to increase election involvement. The common perception among journalists and political Observers is that negative political advertising discourages voters from political participation (Colford, 1988; Hinds, 1990; Martz, 1988; Much Ado, 1985; Denton and Woodward, 1990). In addition, previous research examining the impact of political advertising on election involvement indicates that political advertising has no effect (Garramone, et al., 1990). The findings in the present study, however, indicate that the most negative stimuli actually increase election involvement. While this finding is interesting, it should not be over emphasized. This is only a single indication that negative advertising can increase political involvement. These findings have not been replicated and stand alone against the results of other research, as well 188 as conventional wisdom. In addition, the involvement measures contained in the test instrument are likely to be lacking some degree Of external validity. Asking subjects how stimulating or exciting they find an election does not necessarily translate into greater political interest or participation. Finally, the experimental setting of the research also hinders the external generalizability of the involvement finding. In this instance, subjects are likely to be extremely low in involvement due to a lack Of knowledge, participation, and general interest in the election. On this basis, any additional information concerning the election has the potential to increase involvement. Limitations The limitations of this study derive from a variety of sources including the research method utilized, the independent and dependent variables employed, and the selection of college students as subjects. In a general sense, a laboratory setting is a poor setting in which to conduct political advertising research. While this environment contributes greatly to internal validity, it is a significant hindrance to external validity. In a political campaign, the impact of political advertising takes place among a myriad of communication sources. Competing messages are transmitted via different channels amid a variety Of social and cultural influences. In an experimental setting, each of these influences are 189 necessarily limited in order to examine the relationship between independent and dependent variables. This lack of context hinders the external validity of experimental research. By controlling for extraneous variables, important sources of influence and their interaction with other campaign elements are eliminated. The result is an increase in internal validity at the expense of external validity. A related hindrance to external validity involves the unnatural exposure patterns used in experimental research. Requiring subjects to read biographic profiles and political advertising under tightly controlled time constraints is greatly unlike exposure patterns existing outside of a research laboratory. While attempts were made to approximate a multiple exposure pattern, these attempts fall far short of the exposure patterns existing within a nonresearch environment. The use of a convenience sample of college students also hinders the external generalizability of this study’s findings. While the sample is made up Of subjects of voting age, participants are skewed in terms of both age and education. Many of these subjects arguably lack political sophistication or experience. Roddy and Garramone (1988) suggest that the youth and education of college students ultimately counterbalance, making them adequate subjects for political advertising research. Nevertheless, a convenience sample Of college students hinders the external generalizability of research findings to some degree. 190 In addition, subject responses to test items cannot be accepted as the equivalent of their political behavior or attitudes. While scaled items may be taken as indications of behavior and attitudes, there is a substantial difference between projecting responses onto seven-point scales and the actual behavioral and attitudinal outcomes that are likely to result from exposure to political advertising. In addition, these measures are collected at a single point in time, immediately after exposure to experimental stimuli. Such measurement, while a necessary part of the design of the experiment, may not be accurate in its attitudinal and behavioral representation. As an additional note, while the direction of several scale items was reversed to control for response bias, it is possible that such bias still made its way into subject ratings. In fact, there are a number of potential sources of bias that must be considered in any experimental setting, including the demand characteristics of the research situation and the behavior of researchers. While reasonable controls were employed to contain the introduction of bias, it was simply not feasible to provide all of the controls required to eliminate all potential sources of bias. Future Research Future research examining comparative political advertising should attempt to provide a greater depth Of understanding concerning the findings disclosed in this study. Further information needs to be gathered in order to 191 confirm or disconfirm these results and provide additional information concerning this important field of research. At the most basic level, additional research should be conducted in an attempt to replicate these findings, correcting for the weaknesses existing in this study. Perhaps the most important aspect of such replication should involve the use of a randomly selected sample of the voting public, rather than a convenience sample of college students. Additional improvements in external validity should be obtained by utilizing natural environments and normal media exposure patterns. Another important aspect Of replication involves the use of broadcast media, particularly televised commercials, to study comparative advertising effects. While print media are often used as a vehicle for comparative political advertising, the preponderance of political advertising is presented to the electorate via television. In this instance, an examination of different types of comparative broadcast commercials would provide additional information concerning the role of communication modality in comparative political advertising effects. Finally, research efforts should be made that attempt to separate the influences Of comparative information from the influences of negative information in political advertising. While some social scientists have suggested that comparative advertising is necessarily negative advertising (Johnson-Cartee and Copeland, 1991a, 1991b), it 192 is reasonable to expect that positive comparative political advertising exists, based on the pretest results of experimental stimuli. Such advertising would work to the benefit the sponsor without engendering the perceptions of negativity created by negative political advertising. In this instance, a 2 x 2 factorial design employing comparative and noncomparative advertising, and positive and negative advertising, would be useful in separating the influences of comparative information from the influences of negative information in political advertising. This design would also provide for an examination of interaction effects. Summary Of Conclusions There are several conclusions to be drawn from this study. From a broad conceptual perspective, it is interesting to note that the comparative and noncomparative advertising utilized in this study appear to be perceived in a manner that is fundamentally different. This difference, derived from the negativity of complete comparative advertising, is especially notable in subject ratings Of information credibility. As might be expected, positive noncomparative advertising is rated significantly higher than negative complete comparative advertising in terms of information credibility. When comparative advertising contains some positive information concerning its target, however, the information credibility differences between noncomparative and comparative advertising are nullified. 193 Despite the increased information credibility enjoyed by noncomparative advertising, complete comparative advertising still results in its desired impact. Here, the greater relative influence of negative information over positive information is substantiated. This result is most evident in candidate evaluations. The results Of this study indicate a significant negative change in subject evaluations of the targeted candidate after exposure to negative comparative advertising. These findings provide additional evidence concerning the greater weighting given to negative information over positive information in political perception and information processing. The results of this study provide additional evidence concerning the backlash effect, as well. While evidence Of a backlash effect emerges in both sponsoring candidate evaluations and ratings of sponsoring candidate mean spiritedness, this effect appears to be relatively weak. When comparative advertising contains a high degree of negative information, the results of this study indicate that sponsoring candidate evaluations decline. Sponsoring candidate evaluations also decline if a comparative advertisement does not contain enough information to present a clear rationale for supporting the sponsoring candidate over the targeted candidate. These backlash effects are nullified by moderate comparative advertising, however. In this instance, moderate comparative advertising appears to contain an acceptable range of negative information, nullifying the occurrence of backlash effects and 194 significantly increasing sponsoring candidate evaluations. Additional evidence of a backlash effect is contained in evaluations of sponsoring candidate mean spiritedness. Here, the most negative advertisement, complete comparative advertising, significantly increases perceptions of sponsoring candidate mean spiritedness. It should be noted, however, that perceptions Of mean spiritedness are likely to be important primarily in instances in which they escalate to such an extent that they negatively impact voter behavior. If perceptions of sponsoring candidate mean spiritedness remain within acceptable limits, they are unlikely to harm a candidate. In addition, it is significant to note that this effect is nullified by comparative advertising employing a moderate amount of negative information. Finally, it is of great interest to note the ability of complete and moderate comparative advertising to increase election involvement. While the importance of this finding should not be overstated for reasons already discussed, these results provide preliminary evidence indicating that political advertising can positively impact election involvement. This finding is contrary to the generally accepted notion that negative political advertising creates election apathy and discourages political participation. APPENDIX A EXPERIMENTAL STIMULI COMPLETE COMPARATIVE STIMULUS Check Out DAVE WILSON for 3rd District Senator DAVE WILSON M Is a staunch advocate of increased funding for higher education Brian Rogers Is an opponent of increased funding for higher education Understands the needs and Is out of touch with the needs concerns of local residents and concerns of local residents Bl Strongly opposes the pro- Supports the proposed 4 cent posed 4 cent sales tax increase sales tax increase 8’ Has consistently voted to prevent Has been inconsistent in the the clear-cutting of national protection of national forests forests 8’ Has voluntarily released his Has refused to release his income tax records Has an unquestioned record of honesty in public office income tax records Has had honesty questioned in ethics probe Vote DAVE WILSON for Senator of the 3rd District Paid for by Citizens for Dave Wilson 195 198 NONCOMPARAT IVE STIMULUS Check Out DAVE WILSON for 3rd District Senator DAVE WILSON Is a staunch advocate of increased funding for higher education Understands the needs and concerns V of local residents Strongly opposes the proposed 4 cent sales tax increase Has consistently voted to prevent the clear-cutting of national forests M Has voluntarily released his income tax records [B/ Has an unquestioned record of honesty in public office Vote DAVE WILSON for Senator of the 3rd District Paid for by Citizens for Dave Wilson APPENDIX B INSTRUMENT INSTRUMENT POLITICAL COMMUNICATION STUDY STUDY CONSENT FORM 1 A study is being conducted to learn about perceptions of political candidates and reactions to political campaigns. You will be asked to complete various exercises as part of this study, including: 1. Reading brief biographic profiles of candidates competing for a state Senate seat in Idaho: 2. completing questionnaires concerning your impressions of the political candidates; and, 3. reading political advertisements and/or essays. At the conclusion of the study you will be debriefed regarding the study's theory, methodology, and expected findings. The entire study will take approximately 20 minutes of your time. You may decline to participate or withdraw from the study at any time. All subject responses will be treated with strict confidentiality. Your name and student number are needed only to ensure that you receive proper credit for participating in the study. Your name and/or student number will not be used to identify your individual responses. If you have any questions regarding the study, or would like to obtain a copy of the study’s findings, contact the director of graduate studies in the School of Mass Communications at Texas Tech University. Please indicate your willingness to participate in the study by both signing and printing your name, and writing your student number, in the spaces provided below. PRINTED NAME SIGNATURE STUDENT NUMBER PLEASE WAIT QUIETLY DO NOT TURN THE PAGE UNTIL INSTRUCTED TO DO $0 199 200 IDAHO LEAGUE OF WOMEN’S VOTERS CANDIDATE PROFILES DAVID R. WILSON David R. Wilson, 46, was born and raised in Pocatello, Idaho. Wilson graduated from Boise State University in 1967 with a bachelor of arts degree in criminal justice. After graduation, Wilson returned to Pocatello and started a local retail firm with a family friend. Wilson sold his interest in the business after four years in order to pursue a career in law. Wilson attended law school at Vanderbilt University. He graduated in 1974 with an emphasis in criminal law. He subsequently joined the law offices of Mead, McCaleb, and Grant. Wilson is married with two children. In 1980 Wilson was elected to the Bingham County Board of Directors. He served two terms. Wilson played key roles in initiating new road and bridge construction in Bingham County. As a private citizen, Wilson has held leadership roles with Pocatello chapter of the American Cancer Society and the Physically Impaired Association of Idaho. BRIAN A. ROGERS Brian A. Rogers was born September 17, 1940, in Nampa, Idaho. Rogers enrolled at Idaho State University in 1958. Rogers was a member of the Student Senate for three years. He graduated in 1963 with a bachelor of science degree in political science, prelaw. Rogers attended Wake Forest University Law School. Following graduation, Rogers returned to Nampa where he served as a county attorney for nine years. Rogers is married with three children. Rogers was elected to the Nampa City Council in 1980 and served two terms. Rogers is credited with helping to revitalize downtown Nampa while on the City Council. Rogers is a Red Cross volunteer and has also served on the Board of Directors of the Idaho Council for the Humanities. PLEASE WAIT QUIETLY DO NOT TURN THE PAGE UNTIL INSTRUCTED TO DO SO 201 INSTRUCTIONS: Please complete the following scales. You may refer back to the material on the previous page to help you complete the scales. Please rate Dave Wilson on the following characteristics by circling the appropriate number. Believable 1 2 3 4 5 6 7 ginevable Unconcerned 1 2 3 4 5 6 7 Concerned Dishonest 1 2 3 4 5 6 7 Honest Intelligent 1 2 3 4 5 6 7 Unintelligent Persuasive 1 2 3 4 5 6 7 Unpersuasive Insincere 1 2 3 4 5 6 7 Sincere Unqualified 1 2 3 4 5 6 7 Qualified Good 1 2 3 4 5 6 7 Bad Unethical 1 2 3 4 5 6 7 Ethical Based on the candidate profile, how much do you like Dave Wilson? Like Dislike very much 1 2 3 4 5 6 7 very much How mean spirited is Dave Wilson? Very mean Not at all spirited 1 2 3 4 5 6 7 mean spirited If you were to vote in this election, how likely would you be to vote for Dave Wilson? Very likely 1 2 3 4 5 6 7 Very unlikely 202 Please rate Brian Rogers on the following characteristics by circling the appropriate number. Believable 1 2 3 4 5 6 7 :inevable Unconcerned 1 2 3 4 5 6 7 Concerned Dishonest 1 2 3 4 5 6 7 Honest Intelligent 1 2 3 4 5 6 7 Unintelligent Persuasive 1 2 3 4 5 6 7 Unpersuasive Insincere l 2 3 4 5 6 7 Sincere Unqualified 1 2 3 4 5 6 7 Qualified Good 1 2 3 4 5 6 7 Bad Unethical 1 2 3 4 5 6 7 Ethical Based on the candidate profile, how much do you like Brian Rogers? Like Dislike very much 1 2 3 4 5 6 7 very much How mean spirited is Brian Rogers? Very mean Not at all spirited 1 2 3 4 5 6 7 mean spirited If you were to vote in this election, how likely would you be to vote for Brian Rogers? Very likely I 2 3 4 5 6 7 Very unlikely How interesting do you find the Senate race between Wilson and Rogers? Not at all Very interesting 1 2 3 4 S 6 7 interesting How much do you care about this election? Not at all 1 2 3 4 5 6 7 A lot 203 How stimulating do you find the contest between Wilson and Rogers? Not at all Very stimulating 1 2 3 4 5 6 7 stimulating How exciting is this race? Not at all Very exciting 1 2 3 4 5 6 7 exciting PLEASE WAIT QUIETLY DO NOT TURN THE PAGE UNTIL INSTRUCTED TO DO SO 207 PLEASE TURN THE PAGE AND CAREFULLY READ THE POLITICAL ADVERTISEMENT DO NOT TURN PAST THE POLITICAL ADVERTISEMENT UNTIL INSTRUCTED TO DO SO 205 Check Out DAVE WILSON for 3rd District Senator DAVE WILSON [a] Is a staunch advocate of increased funding for higher education Understands the needs and concerns of local residents E/ Strongly opposes the pro- posed 4 cent sales tax increase g Has consistently voted to prevent the clear-cutting of national forests 8’ Has voluntarily released his income tax records Has an unquestioned record of honesty in public office Brian Rogers Is an opponent of increased funding for higher education Is out of touch with the needs and concerns of local residents Supports the proposed 4 cent sales tax increase Has been inconsistent in the protection of national forests Has refused to release his income tax records Has had honesty questioned in ethics probe Vote DAVE WILSON for Senator of the 3rd District Paid for by Citizens for Dave Wilson 206 INSTRUCTIONS: Suppose that you are discussing the candidates with a friend. Please list the thoughts you have about each of the candidates. Write ONE thought in each box. Use as many boxes as you need. Continue on the back of this page, if necessary. DO NOT continue on to the next page, or turn back to previous pages, until instructed to do so. THOUGHTS ABOUT DAVE WILSON. THOUGHTS ABOUT BRIAN ROGERS PLEASE WAIT QUIETLY DO NOT TURN THE PAGE UNTIL INSTRUCTED TO DO SO 209 INSTRUCTIONS: Please complete the following scales. You may refer back to the material on the previous page to help you complete the scales. Please DO NOT refer to any other portion of the questionnaire. Please rate the political advertisement on the following characteristics by circling the appropriate number. Believable l 2 3 4 5 6 7 ginevable Biased 1 2 3 4 5 6 7 Unbiased Unfair 1 2 3 4 5 6 7 Fair Informative 1 2 3 4 S 6 7 Uninformative Interesting 1 2 3 4 5 6 7 Uninteresting Not useful 1 2 3 4 5 6 7 Useful How negative are the contents of the advertisement you read? Not at all Very negative 1 2 3 4 5 6 7 negative Please rate Dave Wilson on the following characteristics by circling the appropriate number. Believable 1 2 3 4 S 6 7 EZIievable Unconcerned 1 2 3 4 5 6 7 Concerned Dishonest 1 2 3 4 5 6 7 Honest Intelligent 1 2 3 4 5 6 7 Unintelligent Persuasive 1 2 3 4 5 6 7 Unpersuasive Insincere 1 2 3 4 S 6 7 Sincere Unqualified 1 2 3 4 5 6 7 Qualified Good 1 2 3 4 5 6 7 Bad Unethical 1 2 3 4 5 6 7 Ethical 11 210 Based on the political advertisement, how much do you like Dave Wilson? Like Dislike very much 1 2 3 4 5 6 7 _ very much How mean spirited is Dave Wilson? Very mean Not at all Spirited 1 2 3 4 5 6 7 mean spirited If you were to vote in this election, how likely would you be to vote for Dave Wilson? Very likely 1 2 3 4 5 6 7 Very unlikely Please rate Brian Rogers on the following characteristics by circling the appropriate number. Believable 1 2 3 4 S 6 7 :ZIievable Unconcerned 1 2 3 4 5 6 7 Concerned Dishonest l 2 3 4 5 6 7 Honest Intelligent 1 2 3 4 5 6 7 Unintelligent Persuasive 1 2 3 4 5 6 7 Unpersuasive Insincere 1 2 3 4 5 6 7 Sincere Unqualified 1 2 3 4 5 6 7 Qualified Good 1 2 3 4 5 6 7 Bad Unethical 1 2 3 4 5 6 7 Ethical Based on the political advertisement, how much do you like Brian Rogers? Like Dislike very much 1 2 3 4 5 6 7 very much How mean spirited is Brian Rogers? Very mean Not at all spirited 1 2 3 4 5 6 7 mean spirited 12 211 If you were to vote in this election, how likely would you be to vote for Brian Rogers? Very likely I 2 3 4 5 6 7 Very unlikely How interesting do you find the Senate race between Wilson and Rogers? Not at all Very interesting 1 2 3 4 5 6 7 interesting How much do you care about this election? Not at all 1 2 3 4 5 6 7 A lot How stimulating do you find the contest between Wilson and Rogers? Not at all Very stimulating 1 2 3 4 5 6 7 stimulating How exciting is this race? Not at all Very exciting I 2 3 4 5 6 7 exciting PLEASE CLOSE YOUR QUESTIONNAIRE AND WAIT QUIETLY UNTIL GIVEN FURTHER INSTRUCTIONS 13 APPENDIX C BIOGRAPHIC PROFILES BIOGRAPHIC PROFILES IDAHO LEAGUE OF WOMEN’S VOTERS CANDIDATE PROFILES DAVID R. WILSON David R. Wilson, 46, was born and raised in Pocatello, Idaho. Wilson graduated from Boise State University in 1967 with a bachelor of arts degree in criminal justice. After graduation, Wilson returned to Pocatello and started a local retail firm with a family friend. Wilson sold his interest in the business after four years in order to pursue a career in law. Wilson attended law school at Vanderbilt University. He graduated in 1974 with an emphasis in criminal law. He subsequently joined the law offices of Head, McCaleb, and Grant. Wilson is married with two children. In 1980 Wilson was elected to the Binqham County Board of Directors. He served two terms. Wilson played key roles in initiating new road and bridge construction in Bingham County. As a private citizen, Wilson has held leadership roles with Pocatello chapter of the American Cancer Society and the Physically Impaired Association of Idaho. BRIAN A. ROGERS Brian A. Rogers was born September 17, 1940, in Nampa, Idaho. Rogers enrolled at Idaho State University in 1958. Rogers was a member of the Student Senate for three years. He graduated in 1963 with a bachelor of science degree in political science, prelaw. Rogers attended Wake Forest University Law School. Following graduation, Rogers returned to Nampa where he served as a county attorney for nine years. Rogers is married with three children. Rogers was elected to the Nampa City Council in 1980 and served two terms. Rogers is credited with helping to revitalize downtown Nampa while on the City Council. Rogers is a Red Cross volunteer and has also served on the Board of Directors of the Idaho Council for the Humanities. 212 APPENDIX D CONTROL ESSAY l CONTROL ESSAY IDAHO Idaho is one of the group of states known as Mountain States. It is situated in the northwestern part of the United States, bordering the Canadian province of British Columbia. Idaho is roughly L-shaped, with a broad base extending eastward to Wyoming. Idaho measures only about 45 miles from east to west along its northern border. Idaho is approximately 300 miles wide along its southern border. From north to south, along its border with Washington and Oregon, Idaho spans a distance of approximately 500 miles. 4 Idaho includes parts of four of the major geographic provinces of the western United States. The Northern Rockies cover all of central Idaho and most of Idaho’s panhandle. The Middle Rockies lie in a narrow strip along the Idaho-Wyoming border. Adjoining the Middle Rockies is a small triangular section of the Great Basin, extending along most of the Idaho- Utah border. The remainder of the state belongs to the Columbia Plateau, which extends into Idaho from the west. Agriculture, manufacturing based on farm and forest products, tourism and mining form the basis of Idaho’s economy. Among Idaho’s greatest assets are its rich agricultural lands, large tracts of which are still undeveloped. Idaho possess the water resources needed to make such land productive. Expansion of tourism and of industries based in Idaho's farm, forest, and mineral products are expected to bring continued prosperity to Idaho in the future. 213 APPENDIX E CONTRAST WEIGHTING SPECIFICATIONS CONTRAST WEIGHTS Table 29 Contrast weights for information credibility CompingLLimlugiLLml mod w/ non 0 +1 0 -1 = 0 mod w/ comp -1 +1 0 0 = 0 non w/ comp -1 0 0 +1 = o 4 lim w/ non 0 0 +1 -1 = 0 lim w/ comp -1 0 +1 0 = 0 Table 30 Contrast weights for information utility comp w/ lim +1 0 -1 0 = 0 comp w/ non +1 0 0 -1 = 0 lim w/ non 0 0 +1 -1 = 0 comp w/ mod +1 -1 0 0 = 0 mod w/ non 0 +1 0 -1 = 0 214 215 Table 31 Contrast weights for targeted candidate evaluations mueleimitlgullsuul comp w/ lim +1 0 -1 0 = 0 comp w/ non +1 0 O -1 = 0 lim w/ non 0 0 +1 -1 = 0 comp w/ mod +1 -1 O O = 0 mod w/ non 0 +1 0 -1 = 0 Table 32 Contrast weights for targeted candidate liking mmmimm comp w/ lim +1 0 -1 0 = 0 comp w/ non +1 0 0 -1 = 0 lim w/ non 0 0 +1 -1 = 0 comp w/ mod +1 -1 0 0 = 0 mod w/ non 0 +1 0 -1 = 0 Table 33 Contrast weights for targeted candidate voting preferences QQMIMQQLLQIINQLLSEM comp w/ lim +1 0 -1 0 = 0 comp w/ non +1 0 0 -1 = 0 lim w/ non 0 0 +1 —1 = 0 comp w/ mod +1 -1 0 0 = 0 mod w/ non 0 +1 0 -1 = 0 LI ST OF REFERENCES LIST OF REFERENCES Abelson, R. P., Kinder, D. R., and Peters, M. D., and Fiske, S. T. (1982). Affective and semantic components in political person perception. Joo;o_l of Bo;_ooollty oo_ Social Es yohology, $2, 619- 630. Aden. R. C- (1988. May). Telexieeo political acrertieinoi A rerieu of the literature on " ct ." Paper presented at the International Communication Association Convention, New Orleans, LA. I Anderson, N. H. (1965). Averaging vs. adding as a stimulus-combination rule in impression formation. Journal of Experimental Boreholccxl 2o. 394-400. 5» Archibald, S. J. (Ed.). (1971). Tho pollution of politloo. Washington, D.C.: Public Affairs Press. Ash, 8. B., and Wee, C. (1983). Comparative advertising: A review with implications for further research. Adrances in consumer Researchl lo. 370-376- Atkin, C. K. (1981). Mass media information campaign effectiveness. In R. E. Rice and W. J. Paisley (Eds.), Public communication campaigns (pp. 265-279)- Beverly Hills, CA: Sage. Barry, T., and Tremblay, R. L. (1975). Comparative advertising: Perspectives and issues. Journal of Adxcztisingl A, 15-20- Beigel, A. (1973). Resistance to change: Differential effects of favorable and unfavorable initial communications. British Journal of Social and clinical Psycholocxl 12, 153-158. Belch, G. E. (1981). An examination of comparative and noncomparative television commercials: The effects of claim variation and repetition on cognitive response and message acceptance. Journal of Marketing Researcnl 13, 333-349. Benson, T. W. (1981). Another shooting in cowtown. Quarterly Journal of Speechl oz, 347-406. Boddewyn, J. J., and Marton, K. (1978). Qomoorlson adrertieinoi A uorlouioe etucx New York: Hastings House Publishers. 216 217 Boiney, J., and Paletz, D. L. (1991). In search of the model model: Political science versus political advertising perspectives on voter decision making. In F Biccea (Ed. ). Tele_1eion and political acrertieincl uolume l (pp. 3— 25). Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers. Bolster, B. I., and Springbett, B. M. (1961). The reaction of interviewers to favorable and unfavorable information. Journal of Applied Psycholocr1.1a. 97- 104. Briscoe, M. E., Woodyard, H. D., and Shaw, M. E. (1967). Personality impression change as a function of the favorableness of first impressions. Journal or Personalitrl at. 343-357. Byer, W. J., and Cook, E. F. (1985). Comparative advertising's dilemma: How to attack the competition without alienating his customer. Journol or Qonoumer flalkeiingi 2: 67-71- Campbell, D. T., and Stanley, J. C. (1963). Experimental and quasizexnerimental designs IQ: Ififififllgh- BOStOH: Houghton Mifflin. Colford, S. W. (1986, November 10). Pols accentuated negative. aaxertieino Ace pp. 3. 104. Colford, S. W. (1988, October 31). Campaign flak flies: Pols' ads have negative charge. angerrising Age, p. 4. Collyer,. C. E., and Enns, J. T. (1986). Anolyeie Qfi ° The basic deeigne. Chicago: Nelson-Hall Publishers. Crissy, W., and Regan, J. J. (1951). Halo in the employment interview. Journal 91 Applied E§¥9h919921 i5, 338-341. Cusumano, D. R., and Richey, M. H. (1970). Negative salience in impressions of character: Effects of extremeness of stimulus information. Eeyohouomio Sciencel 2o. 81-83. Demirijian, Z. S. (1983). Sales effectiveness of comparative advertising: An experimental field investigation. Journal of consumer Researchl IQ. 362- 364. Denton, R. E., and Woodward, G. C. (1990). Polirioel communication in America. New York: Praeger. 218 Devlin, L. P. (1977). Contrasts in presidential campaign commercials of 1976. Tue genurel Stereo Sneeon Journal1 ea, 238-249. Devlin, L. P. (1982). Contrasts in presidential campaign commercials of 1980. Political Communication Beuienl z, 1- 38. Devlin, L. P. (1987). Contrasts in presidential campaign commercials of 1984. Political communication Eerieul 12. 25- 55. Droge, C., and Darmon, R. (1987). Associative positioning strategies through comparative advertising: Attribute versus overall similarity approaches. Qournol of Marketing Research1 21. 377-388. Earl, R. L., and Pride, W. M. (1980). The effects of advertisement structure, message sidedness, and performance test results on print advertisement informativeness. Journal of Adrertisinol 2. 36-46. Etgar, M., and Goodwin, S. A. (1982). One-sided versus two-sided comparative message appeals for new brand introductions. Journal of consumer Research1 a. 460- 465. Feldman, S. (1966). Motivational aspects of attitudinal elements and their place in cognitive interaction. In S Feldman (Ed. ). cognitire consistencxi Motixational antecedents and benarioral consequences (pp. 75-108). New York: Academic Press. Fiske, S. T. (1980). Attention and weight in person perception: The impact of negative and extreme behavior. Journal of Personalitx and Social Psychologxl 15. 839-906- Garfield, B. (1990, November 5). Let viewers take warning: Political advertising in this country is a travesty. aarertisino Ace. pp. 28-29. Garramone, G. M. (1984). Voter responses to negative political ads. Journalism Quarterlrl .1. 250- 259- Garramone, G. M. (1985). Effects of negative political advertising: The roles of sponsor and rebuttal. Journal of Broadcasting and Electronic Medial 22. 147- 159. Garramone, G. M., Atkin, C. K., Pinkleton, B. E., and Cole, R. C. (1990). Effects of negative political advertising on the political process. Journal or Broadcasting and Electronic Medial Je. 299-311. 219 Golden, L. L. (1976). Consumer reactions to comparative advertising. Advancee in goneumer Research, J, 63-67. Golden, L. L. (1979). Consumer reactions to explicit brand comparisons in advertisements. Journal or Markering Research1 1o. 517- 532. Goodwin, S., and Etgar. M. (1980). An experimental investigation of comparative advertising: Impact of load, and utility of product class. Journal or Marketing Research1 11. 187- 202. Gorn, G. J., and Weinberg, C. B. (1983). Comparative advertising: Some positive results. Aannoee in consumer Research1 19. 377-380. Gorn, G. J., and Weinberg, C. B. (1984). The impact of comparative advertising on perception and attitude: Some positive findings. Journal of consumer Research1 11, 719-727. Gotlieb, J. B., and Sarel, D. (1991). Comparative advertising effectiveness: The role of involvement and source credibility. Journal or Aduerrieingl 29, 38-45. Gray-Little, B. (1973). The Salience of negative information in impression formation among two Danish samples. Journal of crosszcultural Psycholocx1 A. 193-206. Gronbeck, B. E. (1985, November). The rherorio or neoariue political acrertisinc1 Thoughts on senatorial race ads in lgfia. Paper presented at the Speech Communication Association Convention, Denver, CO. Guenther, W. C. (1964). Analysis or yarianoe. Englewood Cliffs, NJ: Prentice-Hall. Hamilton, D. L., and Huffman, L. F. (1971). Generality of impression-formation processes for evaluative and non-evaluative judgments. Journal or Personaliry ano social Psycholocrl 2Q. 200’207- Hamilton, D. L., and Zanna, M. P. (1972). Differential weighting of favorable and unfavorable attributes in impression formation. Journal of Experimental Research in Personalitrl a. 204-212. Harmon, R. R., Razzouk, N. Y., and Stern, B. L. (1983). The information content of comparative magazine advertisements. Journal of Acrertisino1 12. 10-19. Hellweg, S. (1979). An examination of voter conceptualizations of the ideal political candidate. Southern Speech communication Journal1 11. 373- 385. 220 Hill, R. P. (1989). An exploration of voter responses to political advertisements. Journal of Aduertisingi,l§, 14- 22. Hinds, J. (1990, October 29). Negative vibes. perroiu Hens. pp- 1F-2F- Hodges, B. H. (1974). Effect of valence on relative weighting in impression formation. Journal or Personality and Social Psxcholocx1 so. 378-381. Jacob, H. (1971). Black and white perceptions of justice in the city. Lao and Society Berieu1 a. 69-89. Jain, S. C., and Hackleman, E. C. (1978). How effective is comparison advertising for stimulating recall? Journal of Adrertisinc1 2. 20-25. James, K. E., and Hensel, P. J. (1991). Negative advertising: The malicious strain of comparative advertising. Journal of Aduertisinc1 29. 53-69. Johnson, K. S., and Copeland, G. A. (1987, May). Serring the parameters of good taste1 Negatire ool1tical adrertisino and the 12cc elections. Paper presented at the International Communication Association Convention, Montreal, Canada. Johnson-Cartee, K. S., and Copeland, G. A. (1989). Southern voters' reaction to negative political ads in 1986. Journalism Quarterlx1 as. 888-893. 986. Johnson-Cartee, K. 3., and Copeland, G. A. (1991a, August). 3 !' . candidatezsuonsored neoatiue political effects reconsidered. Paper presented at the Association for Education in Journalism and Mass Communication Convention, Boston, MA. Johnson-Cartee, K. S., and Copeland, G. A. (1991bL Negatire political aduertisind1 coming of age Hillsdale, NJ: Lawrence Erlbaum Associates, Publishers. Jordan, N. (1965). The "asymmetry" of "liking" and "disliking:" A phenomenon meriting further replication and research. Puolic opinion.ouarterl¥1 22. 315-322. Joslyn. R. a. (1984). Mass media and elections. Reading. MA: Addison-Wesley Publishing. 221 Joslyn, R. A. (1986). Political advertising and the meaning of elections. In L. L. Kaid, D. Nimmo, and K. R Sanders (Eds. ). Hen perspectixes on political ad_erti§ino (pp. 139-183). Carbondale, IL: Southern Illinois University Press. Kaid, L. L., and Boydston, J. (1987). An experimental study of the effectiveness of negative political advertisements. communication Quarterl¥1 35. 193- 201. Katz, D., Gutek, B. A., Kahn, R. L., and Burton, E. (1975). Bureaucratic encounters. Ann Arbor. MI: Institute for Social Research, University of Michigan. Kellermann, K. (1984). The negativity effect and its implication for initial interaction. communioarion Monographs1451. 37- 55. Kellermann, K. (1989). The negativity effect in interaction: It's all in your point of view. Human communication Research1 1a. 147-183. Keppel. G. (1973). Design and analxs1s1 A researcheris handbook. Englewood Cliffs, NJ: Prentice-Hall. Kernell, S. (1977). Presidential popularity and negative voting: An alternative explanation of the midterm congressional decline of the president's party. The American Political Science Reyieu1 Zl. 44- 66. Kinder, D. R. (1978). Political person perception: The asymmetrical influence of sentiment and choice on perceptions of presidential candidates. Journal of and Social Psxcholoc¥1 as. 859- 871. Lamb, C. W., Pride, W. M., and Pletcher, B. A. (1978). A taxonomy of comparative advertising research. Journal 9: Adyfirfilfiiflgl Z. 43’ 47- Larson, C. U. (1982). Media metaphors: Two models for rhetorically criticizing the political television spot advertisement. central States Speech Journal1 33. 533-546. Lau, R. R. (1982). Negativity in political perception. Political Behauior1 1. 353- 377. Lau, R. R. (1985). Two explanations for negativity effects in political behavior. American Journal of Political Science1 22. 119-138- Lavidge, R. J., and Steiner, G. A. (1961). A model for predictive measurements of advertising effectiveness. Journal or Marketino1 25. 59-62. 222 Lazarsfeld, P. F., Berelson, B., and Gaudet, H. (1948). The peopleis ohoioe. New York: Columbia University Press . Levin, I. P., and Schmidt, C. F. (1969). Sequential effects in impression formation with binary intermittent responding. Journal or Experimental Psycholoor1 12. 233-287- Levine, P. (1976). Commercials that name competing brands. Journal of Adrertisind Research1 lo. 7-14. Louden. A. D. (1987. May). Political adrertisinc in the Huntzflelms.senate race1 Pot shots and hot spots. Paper presented at the International Communication Association Convention, Montreal, Quebec, Canada. Louden. A. D. (1990. June). Transformation of issue to image and presence1 Elicitinc character eualuations in negatire spot adrertisind Paper presented at the International Communication Association Convention, Dublin, Ireland. Martz, L. (1988, October 31). The smear campaign. Neusueek. pp. 16-17- McDougall, G. H. G. (1978). Comparative advertising: The effect of claim type and brand loyalty. current Issues and Research in Adrertisino. 39- 42. Merritt, S. (1984). Negative political advertising: Some empirical findings. Journal of Adrertisinc1 11. 27-38. Meyerowitz, S. A. (1985). Brand X strikes back! The developing law of comparative advertising. Business Marketind1 1Q: 81- 86. Much ado about ads knocking candidates. (1985, September 16). Broadcasting. pp. 64-72. O’Connor, P. J. (1986). The information value of comparative advertising mediating purchase intention. Proceedings of the American Marketing Association Educators. conference1 oz. 23-27. Ohanian, R., and Cunningham, I. C. M. (1987). Application of primacy and recency in comparative advertising. current Issues and Research in Adrertisinc. 99-121. Pechmann, C., and Ratneshwar, S. (1991). The use of comparative advertising for brand positioning: Association versus differentiation. Journal or consumer Research1 1a. 145-160. 223 Perloff, R. M., and Kinsey, D. (1990, August). The consultants look at aduertisino1 Beliefs about political adrertising effects. Paper presented at the Association for Education in Journalism and Mass Communication Convention, Minneapolis, MN. Pfau, M., Kenski, H. C., Nitz, M., and Sorenson, J. (1990). Efficacy of inoculation strategies in promoting resistance to political attack messages: Application to direct mail. communication Monographs1 51. 25-43. Pinkleton, B. E., and Garramone, G. M. (1992, March). A surrey of responses to negatire political adrertisino1 Peter cognition1 affect and,beharior. Paper presented at the American Academy of Advertising Convention, San Antonio, TX. Posch, R. (1986). Comparative advertising. oireor Marketinc1 12 (September). 122-125. Prasad, K. V. (1976). Communication-effectiveness of comparative advertising: A laboratory analysis. Journal of Marketing Research1 13. 128-137. Pride, W. M., Lamb, C. W., and Pletcher, B. A. (1977). The informativeness of comparative advertisements: An empirical investigation. Journal of Aduertisind1 a. 29-35, 48. Richey, M. B., McClelland, L., Shimkunas, A. M. (1967). Relative influence of positive and negative information in impressions and persistence. Journal or Personality and Social Psycholod¥1 a. 322-327- Roddy, B. L., and Garramone, G. M. (1988). Appeals and strategies of negative political advertising. Journal of Broadcasting and Electronic Medial 12. 415-427- Rogers, E. M., and Storey, J. D. (1987). Communication campaigns. In C. R. Berger and S. H. Chafee (Eds. ), Handbook of communication science (pp. 817- 846). Newbury Park, CA: Sage. Rogers, J. C., and Williams, T. G. (1989). Comparative advertising effectiveness: Practitioners' perceptions versus academic research findings. Journal or Adrertisind Research1 29. 22-37. Rosenthal, R., and Rosnow, R. L. (1985). gonrrasr anal¥s1s1 focused comparisons in the analxsis of uarianoe. New York: Cambridge University Press. Rothschild, M. L., and Ray, M. L. (1974). Involvement and political advertising effect: An exploratory experiment. Communication Research, l, 264-284. 224 SabatO. L. (1981). The rise of political consultants. New York: Basic Books. Salmore, B. G., and Salmore, S. A. (1989). Candidares, parriesr ano campaign . Washington, D.C.: Congressional Quarterly Press. Scammon, D. (1978). Comparative advertising: A re- examination of the issues. The Journal or gonsumer Affairs1 12. 381-392. Shapiro, M. A., and Rieger R. H. (1989, May). Qomparing positire and negati_e political adrertisinc. Paper presented at the International Communication Association Convention, San Francisco, CA. Shegula, D. A., and Jacoby, J. (1978). Do comparative claims encourage comparison shopping? ourrenr Issues and Research in Adrertisinc. 23-37. Shimp, T. A., and Dyer, D. C. (1978). The effects of comparative advertising mediated by market position of sponsoring brand. Journal of Aduertisinc1 1. 13-19. Shyles, L. (1986). The televised political spot advertisement: Its structure, content, and role in the political system. In L. L. Kaid, D. Nimmo, and K. R. Sanders (Eds. ). Hen perspectiues on political aouerpising (pp. 107- 138). Carbondale, IL: Southern Illinois University Press. Snedecor, G. W., and Cochran, W. G. (1967). srarisrioal meuhoos. Ames, IA: Iowa State University Press. Springbett, B. M. (1958). Factors affecting the final decision in the employment interview. Qanaaian Journal of Psycholoc¥1 12. 13-22. Stack. R. T. (1978). comparatire adxertisino1 A determination of effects. Unpublished doctoral dissertation, Michigan State University, East Lansing. Steward, C. J. (1975). Voter perception of mudslinging in political communication. Central States fincccn Journal1 2a. 279-286- Stutts, M. (1982). Comparative advertising and ccunterargument. Journal of the Academr of Marketing Science1 1o. 302-313- Sujan, M., and Dekleva, C. (1987). Product categorization and inference making: Some implications for comparative advertising. Journal of Consumer Research1 1g, 372-378. 225 Surlin, S. H., and Gordon, T. F. (1977). How values affect attitudes toward direct reference political advertising. Journalism Quarterly1 as. 89-95. Swayne, L. E. and Stevenson, T. H. (1987). Comparative advertising in horizontal business publications. Industrial Marketing Management1 it. 71-76. Swinyard, W. R. (1981). The interaction between comparative advertising and copy claim variation. Journal of Marketing Research1 18. 175-186. Tashchian, R. O., and Slama, M. E. (1984). Involvement and the effectiveness of comparative advertising. Qurrenr Issues and Research in Aduertisinc. 79-92. Trent, J. S., and Friedenberg, R. V. (1991). Polirioal campaign communication1 Principles and practices. New York: Praeger. Turgeon, N., and Barnaby, D. J. (1988). Comparative advertising: Two decades of research. currenr Issues and Research in Adrertisinc. 41-65. Villarreal-Camacho, A. (1985). Effects of product class knowledge on the evaluation of comparative versus noncomparative messages. Aduanoes in consumer Research1 12. 504-509. Wadsworth, A. J., and Kaid, L. L. (1987, May). lnoumhent and challenger styles in presidential aduertisino- Paper presented at the International Communication Association Convention, Montreal, Canada. Walker, D., Richardson, R., Williams, 0., Denyer, T., and McGaughey, S. (1972). Contact and support: An empirical assessment of public attitudes toward the police and the courts. North Carolina Lam Rerieu1 S1. 43-79. Walker, B. A., Swasy, J. L., Rethans, A. J. (1986). The impact of comparative advertising on perception formation in new product introduction. Aguanoes in consumer Research1 13. 121-125. Warr, P., and Jackson, P. (1976). Three weighting criteria in impression formation. European Journal of Social Psxcholodr1 a. 41-49. Webster. E. C- (1964). Decision making in the employment inreruie_. Montreal: McGill University Relations Center. 226 Williams, T. G. (1978). An experiment in the effects of comparative advertising on brand awareness, comprehension and preference. Southern Business Reuieu, A, 30-42. Wilson, R. D. (1976). An empirical evaluation of comparative advertising messages: Subjects' responses on perceptual dimensions. Aguanoes in gonsumer Research1 3. 53-57. Wilson, R. D. (1978). Comparative advertising: Some current considerations for managerial planning and strategy. Current Issues and Research in Adrertis1nd. 5-22. Wilson, R. D., and Muderrisoglu, A. (1979). An analysis of cognitive responses to comparative advertising. Aduances in Consumer Research1 Z. 566-571. Winer. 3- J. (1971). Statistical principles in experimental design. New York: McGraw—Hill. Wyer, R. S. (1970). Information redundancy, inconsistency, and novelty and their role in impression formation. Journal of Experimental Social Psycholoc¥1 6. 111-127.