MEMORABLE POLITICAL MESSAGES: RETHINKING THE ROLE OF CONVERSATION IN PUBLIC OPINION By Hillary Cortney Shulman A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Communication 2011 ABSTRACT MEMORABLE POLITICAL MESSAGES: RETHINKING THE ROLE OF CONVERSATION IN PUBLIC OPINION By Hillary Cortney Shulman The focus of this dissertation research was to investigate students’ memorable political messages. The memorable messages literature (Knapp, Stohl, & Reardon, 1981) has demonstrated that within domains other than politics, people have recalled messages that serve as a guide for their beliefs, attitudes, or behavior within that domain. This study extended this logic to the political realm in an effort to identify the content of political messages reported as memorable. By understanding what types of political messages people identify as being memorable or important, and through knowing characteristics about the source of these messages, more can be known regarding student’s political development. Guidance from Burleson’s (2009) message centered approach to interpersonal communication was used to deduce whether certain types of messages, coming from certain types of sources, were more likely to come from politically engaged participants. Thus, relationships between message substance, message source characteristics, and political engagement variables were also tested. In order to obtain students’ memorable message, 191 participants were solicited from communication courses and asked to complete an online survey regarding their political opinions. After memorable messages were explained, students were asked to record their political memorable message if one existed. After students reported their message, several follow-up questions assessed including measures of message valance, the identity and role of the source, the channel through which the message was transmitted, and the perceived credibility and similarity of the source. Classification of memorable political messages was accomplished through an author-generated system and though a card sorting task (additional N = ??) and multidimensional scaling. This method led to the creation of a classification scheme for the content of memorable messages and tested whether message and source characteristics associated with political outcomes of interest including: scores on a political knowledge test, a political participation index, self-reported political interest, and an index created from these three measures representing a general engagement index. Results from this dissertation research suggest: a) a vast majority (98%) of students report having a memorable political message, b) the content of these memorable messages can be classified into ten content areas, c) students are more likely to report positive messages that originate from the mass media, and d) students who have the highest levels of political engagement tend to be those reporting a negative message from a non-credible source, or those reporting a positive message from a highly credible source. Although the latter relationship was predicted, the finding that negative messages were consistently reported from those interested and engaged in politics has several practical and theoretical implications discussed through the course of this manuscript. ACKNOWLEDGMENTS During my six years in East Lansing, I have developed a long list of people to thank for both helping me complete my degree and for making my time here so enjoyable. I admit that when I first arrived in East Lansing, I was not excited about the social prospects this area afforded. However, soon thereafter, upon meeting all my future mentors and friends, I instantly knew I made the right decision. First off, I owe a debt of gratitude to my guidance committee. Especially Dr. Tim Levine, my advisor for 6 years, with whom I shared both victories and defeats. What I most appreciate about our relationship is that regardless of the occasion, and what challenges I faced, we were always able to laugh about it over drinks. With you I learned the importance of thinking outside the box, challenging the status quo, and embracing creativity even when others do not readily embrace these ideas. To my committee: Dr. Dan Bergan, Dr. Maria Lapinski, and Dr. William Jacoby, I thank you for caring enough about my projects and my own scholarly development to be critical of my work and willing to reread several iterations of my papers. At the end of my graduate school career, not only do I have projects that I am proud of but I have also been equipped with ample stories of my arduous times in school to share with my future advisees. Aside from the formal mentorship from my guidance committee, I also must acknowledge the informal mentorship I was lucky enough to receive. Dr. Gwen Wittenbaum, Dr. Joe Walther, Dr. Chuck Atkin, Dr. Sandi Smith, Dr. Frank Boster, Dr. Kami Silk, and Marge Barkman all contributed to my education in so many ways. From encouraging me to pursue independent projects, to writing letters of recommendation, and to just being good friends who share good advice, I thank all of you. iv My friends have made my graduate experience truly amazing. Particularly the members of my numerous softball teams and the Landsharks women’s basketball team. I would also like to specifically thank my officemate for four years and friend Dr. Lindsay Neuberger (“this is too good to break up”), and Dr. Dave DeAndrea, my best friend. I could not have done this without you…seriously. And finally I must thank my incredible family. Dad, mom, Mandi, I love you all and so appreciate your willingness to support me in every way possible my entire life. I can’t wait to finally rejoin you back in Chicago. v TABLE OF CONTENTS LIST OF TABLES viii LIST OF FIGURES ix CHAPTER 1 INTRODUCTION Definitions and Measurement The Substance of Political Conversations The Added Utility of Context The Impact of Memorable Messages on Civic Engagement - Identifying Outcomes CHAPTER 2 METHOD Overview Participants Survey Procedure Card Sorting Task Procedure Independent Variables Memorable Messages Memorable Message Content Coding Message Valance Message Source Source Credibility Perceptions of Similarity Outcome Variables CHAPTER 3 RESULTS Research Question One Media-Related Messaging Attitudinal Messaging Political Belief Messages Political Experiences Research Question Two Research Question Three Research Question Four Research Question Five Hypothesis One Hypothesis Two Hypothesis Three 1 3 6 11 14 18 18 19 19 19 20 20 22 24 24 25 26 27 28 28 29 29 31 31 32 32 33 34 34 35 vi CHAPTER 4 DISCUSSION Discussion of Results Methodological and Theoretical Insight Limitations and Future Directions Conclusion 40 41 47 56 59 APPENDICES 63 REFERENCES 83 vii LIST OF TABLES Researcher generated coding scheme categories compared with card sorting coding categories, along with number of messages falling into each category and number of interpersonal versus mass communicated messages. 71 Civic engagement means by all memorable message categories. 73 Predicting civic engagement by message valance and source credibility. 74 Correlations between message valance and civic engagement variables under conditions of low, medium, and high source expertise. 75 Multiple regression equations examining whether message valance, source credibility, and source similarity predict levels of political engagement. 76 Correlations between message valance and political engagement variables under conditions of low, medium, and high source credibility and similarity. 77 viii LIST OF FIGURES Mean plot portraying the influence of message valance and source credibility on participants’ political interest. 79 Mean plot portraying the influence of message valance and source credibility on engagement index scores. 80 A visual depiction of the three dimensional solution obtained using multidimensional caling with ALSCAL. 81 ix Introduction “Interpersonal relationships seem to be “anchorage” points for individual opinions, attitudes, habits and values. That is, interacting individuals seem collectively and continuously to generate and to maintain common ideas and behavior patterns which they are reluctant to surrender or to modify unilaterally” (Katz & Lazarsfeld, 1955, p. 44). Talking about politics is commonplace and consequential. According to the National Election Studies, 92% of Americans report having political discussions with their peers (Delli Carpini, Jacobs, & Cook, 2007). A network analysis undertaken by Huckfeldt and Sprague (1991) revealed that people tend to discuss politics within their immediate social network, which tends to be politically homogenous. Klofstad’s (2009) panel study with incoming freshman reported that frequency of political discussion with roommates positively predicted participation in political and organizational activities around campus. Furthermore, a research review by Delli Carpini et al., (2007) reported that frequency of political conversation is positively associated with ideological sophistication, sustained political activity, and political efficacy. For these reasons and more Kim, Wyatt, and Katz (1999) assert that, “conversation is the soul of democracy,” because through informal political talk, “…citizens can bridge their personal experiences with the political worlds out there” (p. 362). Taken together, much research documents the important role of political talk in public opinion and political activity. Work in this vein merits further attention. The current research is interested in how interpersonal communication about political topics impact civic engagement. Specifically, this research examines the impact of memorable messages on civic engagement from the perspective of Burleson’s (2009) message centered approach to interpersonal communication. 1 The idea that political conversation with peers plays a seminal role in the transmission of political beliefs is not new. Katz and Lazarsfeld (1955) made the argument that although mass media likely provides material for political conversations, the true mechanism underlying political influence is interpersonal in nature. When their research was conducted, the field of communication science was still in its nascent years. Currently, however, theoretical and methodological advances in the area of interpersonal communication provide the opportunity to reinvigorate the ideas of Katz and Lazarsfeld. In this study tenets from Burleson’s (2009) message centered approach to interpersonal communication will provide a conceptual framework to empirically test the influence of interpersonal political messages on public opinion. In Burleson’s (2009) message centered perspective, understanding the influence of messages requires knowledge of manner, substance, and outcomes. Manner refers to features of a message aside from the message itself and includes characteristics such as source, context, and relationship between conversational partners. Substance refers to the message content and can be measured in terms of topic, valance, or intended function. Finally, of particular interest here is what effect manner and substance have on specified outcomes. Using this framework to guide the current research, this paper investigates how memorable political messages (substance) with others (manner) influences civic engagement (outcome). Memorable messages are “verbal messages that are remembered for a long period of time and which people perceive as a major influence on the course of their lives” (Knapp, et al., 1981, p. 27). Memorable messages have, by definition, an important influence on individuals. Therefore, memorable messages seem well suited to study the impact of interpersonal communication about politics. Accordingly, this research applies the concept and method of memorable messages to politics and civic engagement. 2 To gain insight into the relationship between interpersonal communication and civic engagement, this paper will dissect the manner, substance, and outcome components of Burleson’s model as applied to memorable political messages. This undertaking begins with a brief review regarding the measurement of political talk. Pursuing this further, the next section suggests an alternative to prior measurement by proposing a memorable message approach to assess message substance. To contextualize this information, the third section focuses on different sources of political messages. In the final section, considerations regarding manner and substance are jointly considered to determine whether certain message features can be attributed to the promotion, or suppression, of desirable political outcomes. In this study, the outcome of interest is civic engagement. Definitions and Measurement Casual political discussion between citizens is a vital aspect of a functioning democracy. Not only do people transmit relevant instrumental information such as polling place details, but also Katz and Lazarsfeld (1955) argued that people seek out opinions from peers who they believe to be credible purveyors of political information. Thus the practical costs of information seeking in a media rich environment can be simplified through interpersonal networks (Huckfeldt & Sprague, 1991; Katz & Lazarsfeld, 1955; Robinson 1976). Due to the importance, ease, and functional versatility of interpersonal political talk, the valid measurement of this construct is paramount. Definition and measurement considerations are therefore the focus of this section. Interpersonal political conversations are exchanges that occur between people with an established relationship about politics. Although this definition provides a starting point for this investigation, admittedly, defining interpersonal exchanges and politics is anything but 3 straightforward. This problem of definition was one of the issues that launched this project. To deal with this, first the issue of a non-unified understanding of the term politics will be discussed. Defining interpersonal exchanges will be the focus of subsequent sections. “The personal is political” (Hanisch, 1969) is a phrase that encompasses the complexity of the term politics. Politics, broadly defined, can be any issue that has implications for public policy. Wyatt, Katz, and Kim (2000) sought to address this definitional concern by listing nine different topics people discuss (e.g., sports/entertainment, personal issues, the economy, etc.), and then through exploratory factor analysis clustered these topic categories into three facets: the political facet, bridge facet, and personal facet. Included in the political facet were conversations about national government, local government, foreign affairs, and the economy. The personal topic list included: religion, sports, and personal issues. Finally, the bridging facet consisted of items that fell into both categories and included crime and education. At face value, these distinctions seem reasonable, however upon closer inspection, there are concerns with these categorizations. The first concern revealed later in the study was that political talk was found to commonly occur in places of worship (Wyatt, et al., 2000). Thus, although people considered religion personal, they were using this space to transmit political beliefs. Additionally, consider the hot button issues of abortion and gay marriage. These issues can be considered highly personal and political in nature, especially for people who are directly affected by this legislation. Wyatt et al. provides an example of an inherent difficulty with measuring and defining political topics – within any context, almost any topic can be considered political or personal. This lack of consensus regarding what constitutes a political conversation is troublesome given that the most prevalent measure of political talk requests that participants approximate the 4 amount of, or frequency of their political discussions (e.g., Delli Carpini et al., 2007; Kim et al., 1999; Klofstad, 2009; Mutz, 2002, 2006; Price, Nir, & Capella, 2002; Wyatt et al., 2000). This measure of political talk has led to the documentation of several important relationships between increases in political conversation and a host of politically relevant outcomes such as: political knowledge, political sophistication/crystallization, increases in political behavior, and increases in political efficacy (for a review see Delli Carpini et al. 2007). Although these relationships demonstrate beneficial outcomes associated with political talk, these studies relied on measures assessing how often people talk about politics. Given the nebulous nature of the term ‘politics’, understanding has not been gained regarding the substance of these conversations. Further, from a survey construction standpoint, vague terms inflate measurement error associated with subjective interpretations of non-primitive terms. Therefore, the relationships reported from this line of research might be stronger, or more nuanced, or both than currently thought. The second commonly used approach to measuring political discussions is to actually have participants engage in a political discussion. This can be done experimentally, through the formation of ad-hoc groups that come into the laboratory and discuss politics under certain conditions (e.g., Druckman & Nelson, 2003; Mackie, 2002; Shulman & Wittenbaum, 2010; Sunstein, 2002). The utility of these experiments is in the ability to isolate one or two variables of interest in order to reveal causal relationships. A related commonly used method to promote political talk is to recruit a random sample of participants to attend a discussion that focuses on public affairs (e.g., Deliberative Democracy Movement, Fishkin & Luskin, 1999). Discussion attendees are usually provided with pre-discussion materials, led by a trained moderator during the actual discussion, and compensated for their participation. Though both of these approaches are important in that they: a) systematically identify the impact of group discussion on beliefs, b) 5 provide people with the opportunity to discuss political affairs in an organized and diverse setting, and/or c) allow discussions to be observed; one important limitation is ubiquitous. Studies that randomly assign participants to condition or assemble nationally representative groups suffer from the absence of ecological validity. Ordinary political conversations do not occur in this manner. In contrast to experiments and observational studies political conversations usually take place with known others (Conover & Searing, 2005; Conover, Searing, & Crewe, 2002; Huckfeldt & Sprague, 1991; Mutz, 2002, 2006; Robinson, 1976), occur spontaneously without prior preparation (Conover et al., 2002; Conover & Searing, 2005) and are initiated for social and personal reasons (Chambers, 2003; Conover & Searing, 2005; Kim et al., 1999). Thus, although randomly assigning people to participate in ad-hoc groups has its scientific virtues, this method cannot be considered an allegory to political conversations as they naturally occur. In order to address these concerns, the next section proposes a way to begin measuring how people think about politics, and is guided by research in memorable messages. Asking participants to recall political messages, and creating a working typology that emerges from these responses can inform how this term is conceptualized in the population. Furthermore, building a taxonomy of political content can be used as a tool for future research concerned with how differential messages lead to diverse outcomes. The Substance of Political Conversations Utilizing Burleson’s (2009) ideas of manner, substance, and outcomes, the purpose of this section is to provide an approach to measure the substance of political messages. An existing program of research that emphasizes the importance of message content is work done in memorable messages (Knapp et al., 1981). Underlying all memorable message research is the belief that people make sense of the world through messages that have been communicated to 6 them. A message here is defined as a receivers’ perception that a symbolic exchange of meaning has been transmitted from another person. This approach to communication research is novel because rather than presenting stimulus materials and recording reactions, this work takes a step back and asks the population of interest to recall important and memorable messages in a domain of interest. Thus, this perspective contributes to past work by looking at the content of political messages rather than the frequency of occurrence, and by recording conversations that occurred under natural circumstances. By applying this paradigm to the problem of political measurement, a firmer understanding regarding what people talk about when they talk about politics can be achieved. In their seminal article Knapp et al. (1981) defined memorable messages as, “verbal messages that are remembered for a long period of time and which people perceive as a major influence on the course of their lives” (p. 27). Knapp et al. reported that people tended to recall messages that were short, interpersonally transmitted, direct, action-oriented, rule focused, and tended to be sent earlier in life. The methodology used to obtain memorable messages is relatively consistent across the literature. Participants are asked to recall a message concerning an area of interest, and are given a few examples of what these messages might look like to help prompt ideas for participants. Participants then write down these messages and answer a series of questions pertaining to the context including when these messages occur, from whom, and why the message was memorable. In the original article on memorable messages, Knapp et al. found that even when they asked the same participants three months later about their memorable message, all participants accurately recalled the content of their message with almost no mistakes. This follow-up sheds light on the resiliency of these messages and is consistent with the validity of the memorable message construct. 7 Memorable message research has revealed that people’s memorable messages provide insight into a person’s beliefs, attitudes, and behaviors within a given domain. For example, Stohl (1986), interested in organizational socialization, found that all workers (100% of the sample) could recollect a key message they were told during their first days in the office. Furthermore, this message served as a salient guideline for behavioral decisions within the office for new workers concerned with easing through the assimilation process. Another area where memorable messages have been explanatory is in family socialization (Medved, Brogan, McClanahan, Morris, Shepherd, 2006). Medved et al. examined whether females and males were socialized with different expectations regarding career expectations, career choice, and family obligations. In order to assess socialization, Medved et al. collected over 900 memorable messages related to work, family, and balance. An analysis of these messages revealed that although women and men both received similar messages regarding the importance of a career and family, only women recalled messages about choosing jobs that would facilitate family planning. This work suggests that because women tend to recall messages alluding to the permeable nature of their career, women tend to make different – and less secure – vocational decisions than men. In the political realm, it stands to reason that political memorable messages also exist. The work cited above was from organizational communication and family communication. In addition to these areas, memorable message research has also been used to examine work ethic in athletes (Kassings & Pappas, 2007), attitudes toward aging (Holladay, 2002), self-assessment and values (Smith & Ellis, 2001; Smith, Ellis, & Yoon, 2001; Ellis & Smith, 2004), health behaviors and support (Ford & Ellis, 1998; Smith et al., 2009), and political socialization (Shulman, Smith, & Clark-Hitt, in progress). Given the diverse range of scholarship that has 8 utilized this method and garnered insight into the content of influential messages, it seems likely that these benefits will extend to the political realm. Thus far, the necessity for deciphering the content of memorable political messages has been argued from different vantage points. The first argument was that the content of communication is important but neglected by research programs where the mere frequency of communication is measured. By allowing subjective interpretations of the term “politics” inform this line of research, clear relationships are obfuscated. Further, it is unlikely that all political talk, regardless of subtopic or context, is equally influential. Instead, some talk is likely to have more impact than other talk. Thus, honing in on what talk is influential and what people think about when they think about politics allows for the refinement of measurement. A second argument presented is that researchers make implicit assumptions about the communication that occurs in their area of interest. The assumption made in political communication research is that people are talking about ‘substantive’ political matters with others because political talk positively correlates with desirable political outcomes. However, without understanding what people are talking about, this relationship might be distorted due to people who are having conversations that are low in quality and tangentially related to the political process. Wyatt et al.’s (2000) research provides an instructive example. Although people report discussing political topics such as National Government issues, there is still little knowledge gained about the substance of these conversations. For example, are people talking about the national unemployment rate, their general distaste for the President, election predictions in different States, or what designer the First Lady was wearing in her last public appearance? By understanding what people recall within a domain of interest, behaviors and attitudes in that context can be better understood. Accordingly, presuming that people have, and 9 can report, memorable political messages, the first objective of this research is to classify memorable messages about politics. RQ1: What types of content are recalled in people’s memorable political messages? Aside from assessing the subject matter present in people’s memorable political messages, message substance can also pertain to message valance. Message valance refers to the general attitude espoused in the message and can range from positive (i.e., favorable, optimistic, encouraging) to negative (i.e., unfavorable, pessimistic, or discouraging). Message valance is considered here because in the political world, people broadly interpret political events, figures, and/or policy as being positive or negative. Furthermore, people can take a positive approach to politics (e.g., politics is important) or a negative approach to political matters (e.g., politicians are corrupt, never get anything done, etc). Therefore, aside from ascertaining what topics people recall in their memorable political message, it is also necessary to address whether people perceive their message as being generally positive or negative. Research on message valance has demonstrated that positive versus negatively framed messages yield different effects on message recipients. For example, Robberson and Rogers (1988) examined whether positive health messages were more persuasive than negative health messages. Positive messages emphasized the positive consequences associated with adherence to the message recommendation, whereas negative messages focused on threats associated with failing to follow the prescribed behaviors. This study revealed that the threat of danger was more persuasive than the promise of good health. Thus, message valance affected persuasiveness. In the political realm, message valance has generated a proliferate line of scholarship related to the effects of ‘positive’ versus ‘negative’ campaigning (for a meta-analysis see Lau, Sigelman, Heldman, & Babbitt, 1999). Positive campaigning is generally considered campaigns 10 that focus on the positive qualities and credentials a candidate can offer the electorate, along with their opinions on issues. Conversely, negative campaigns operate by admonishing the other side’s candidate through accusations of misdeeds. Through this mudslinging the candidate with less skeletons in their closet prevails. The conclusion regarding the campaign strategy debate does not have a simple answer and is beyond the scope of this paper. What is germane to the current investigation, however, is that message valance influences receivers of the message in tangible ways. Thus, to better understand message substance, an understanding of participant’s perception of message valance should also be considered for a more comprehensive analysis. A research question related to the issue of valance is here advanced. RQ2: Do people tend to recall more positive or negative memorable political messages? The Added Utility of Context Although knowing the substance of peoples’ political conversations has merit and potential to fuel future research, taking messages out of context can lead to serious misinterpretations. A compelling series of experiments by Cohen (2003) demonstrated this fact when comparing the relative influence of message content and source cues on political preferences. Specifically, Cohen brought self-identified partisans into a lab where they read a political argument. Participants were randomly assigned to receive either a liberal or conservative argument, with a correct or incorrect source cue (party consistent-party inconsistent), or no source cue at all. In the no source condition, partisans evaluated the message according to their loyalties. When a source cue was present, however, participants supported messages coming from their own political party and were contrary to messages from the opposing party regardless of message content. The Cohen study documents that knowledge of 11 message content alone does not tell the complete story. Outside the lab it is almost impossible to separate the source of the message from the message content. Therefore, thorough understanding of political messages requires an investigation into the manner in which the message was communicated. This requires an investigation into the source, channel, and relationship with the source of the message. In memorable message research, the importance of identifying the message source is pivotal. In most studies, participants’ memorable messages are solicited, with follow up questions inquiring about the source of this information. Previous research has revealed that different sources are often accredited for different types of information, and also different sources yield differential effects on the receiver. For example, in Smith et al.’s (2009) study on memorable messages in the context of breast cancer (N=359), the research found that although many recalled messages were from the media (36%), only messages from medical professionals were positively related to participants adopting appropriate breast cancer detection behaviors. Messages from all other sources (family, friends, media, and others) were not significantly associated with detection or prevention behaviors. This research demonstrates that the interaction between message content and source is fundamental to understanding how these messages influence behavior. In the political realm, identifying where people get their political information from is of great consequence. Research dating back to two-step flow (Katz & Lazarsfeld, 1955) argued that the media provides political information, audience members (i.e., opinion leaders) who are particularly attentive to politics consider this information, and then take it upon themselves to filter media news to “less attentive” audiences. Although Robinson (1976) questions the causal direction the two-step flow model purports, this research verifies that people listen to others who 12 they perceive to be politically knowledgeable. Thus, in an effort to revitalize these ideas, this project reexamines whom people cite as the source of memorable political messages. In order to assess message source, two characteristics will be examined: the source’s role, and whether the message was transmitted interpersonally or from the mass media. Source role refers to how participants describe the source. This can mean participant’s relationship to the source (parent, sibling, friend, teacher), or can refer to occupation (news anchor, talk show host, political candidate). Additionally, of importance is whether the message was transmitted interpersonally or through the mass media. Up to this point, interpersonal conversations have been the focus of this paper. It is acknowledged, however, that the mass media does play a large role in disseminating political information. Consequently, there will be participants that recall a memorable message that was transmitted through the mass media, and was uttered by a politician, pundit, celebrity or so on. It now becomes necessary to define what is meant by interpersonal communication and mass communication, and to further clarify how message source and message channel are being used here. Interpersonal communication refers to a discussion with a person with whom there is an established and reciprocal relationship. Thus, interpersonal communication requires that the people communicating mutually know one another beyond just one superficial encounter. This definition can be considered a relaxed version of Miller and Steinberg’s (1975) definition of interpersonal communication, which requires people know one another at a psychological level. Thus, a source will be considered interpersonal in nature if the message source and participant share a mutually acknowledged relationship for an enduring period of time. Channel traditionally refers to the medium through which the message was transmitted. However, for this project, the medium (i.e., phone, face-to-face, computer mediated, television, 13 etc) is less important than the level of message personalization implied by the medium. A mass communicated message is a message transmitted to a large number of unknown others. Examples of mass communication include a televised speech, a rally speech, a radio or television show, etc. On the other hand, regardless of the medium through which the message was sent, an interpersonal message is more personalized than a mass message. This message is intended for a very specific and small audience, is highly tailored, and transactional in nature meaning immediate feedback is possible and expected. Taken together, these definitions provide guidance for how interpersonal and mass communication sources will be classified. Of interest is the source of memorable political messages, and whether different types of content are particular to certain types of sources. RQ3: Which types of sources, in terms of role and channel, are more likely to be recalled? RQ4: Does the content of memorable political messages differ by source? This section argued that knowledge regarding the source of memorable political messages yields valuable insight regarding whom people cite as being politically influential. Additionally, a distinction was made between interpersonal and mass media sources. The next section amalgamates the ideas discussed so far to investigate whether these source and message characteristics translate into tangible political outcomes. The Impact of Memorable Messages on Civic Engagement – Identifying Outcomes In a democratic society, it is imperative that citizens are provided with an opportunity to arrive at their political preferences and participate according to these preferences. As it stands, in general citizens’ ability to cogently arrive at political conclusions is often disappointing (Converse, 1964; Delli Carpini & Keeter, 1997). Furthermore, the frequency of the publics’ 14 political participation is rather low (Downs, 1957). This being the case, this paper argues that because people are allowed to vote and are expected to participate, scholarly attention should gravitate toward improving this process. Therefore, the final objective of this research is to better understand the content, manner, and substance of messages that yield positive democratic outcomes on their recipients. Positive democratic outcomes are here referred to as political (or civic) engagement, which encompasses the idea that people should be a) knowledgeable about politics, b) interested in politics, and c) participate actively in political affairs. Public opinion scholars are concerned with the quality of opinions. Although opinion quality has been measured in a variety of ways including sophistication (Rosenberg, 1988; Tetlock, 1986), consistency or constraint (Converse, 1964; Zaller & Feldman, 1992), stability (Converse, 1964), and knowledge of political facts (Delli Carpini & Keeter, 1997; Zaller, 1992), the pursuit of this line of research is to understand when opinion data provide veridical feedback for elections and for policymakers (e.g., Stimson, 2004). Despite a lack of consensus regarding what constitutes a high quality opinion, Zaller (1992) argued that knowledge of political facts, and interest in the political system provide the most valid proxies when measuring this construct. On the other hand, literature focusing on political participation examines how people can be mobilized to participate in political activities and why participation occurs (e.g., Downs, 1957; Fowler, 2005; Nickerson, 2008; Verba, Schlozman, & Brady, 1995). Although theoretical models differ regarding how citizenship develops (attitudes then behavior versus behavior motivating attitudes), years of research proffers that holding high quality political beliefs, and exercising these beliefs through participation, are valued practices for sustaining democracy. The question of interest is whether communication features reflected in memorable political messages can be identified that relate to civic engagement. The communication features 15 discussed so far include message content, valance, and source. In terms of message content, the possibility that certain types of messages will correspond with different levels of political engagement provides an interesting starting point. Because the message content classification scheme has not been created, it is not possible to proffer hypotheses regarding what content will be associated with higher levels of civic engagement. Hence, this research question takes an exploratory look at whether message content systematically relates to levels of civic engagement. RQ5: Are levels of civic engagement similar across individuals recalling different message content categories? One reason why levels of civic engagement might differ across content types is because message content can be positive or negative. This is when consideration of message valance provides further explanatory power. It is possible that people who recall negative political messages display low levels of civic engagement. Research on negative campaigning has documented that a negative political climate disengages some voters due to their disenchantment with the political process (Ansolabehere & Iyengar, 1995). Alternatively, positive messages regarding political affairs might provide a catalyst for participation. Chambers (2003) as well as Verba et al. (1995) found that positive political experiences enhance the potential for future participation. Thus, people who recall a positive political message off the top of their head may be people who have had largely positive experiences with politics. These positive experiences should promote enduring levels of political participation, knowledge, and interest. Given these relationships have been found in related literatures, the first hypothesis is advanced regarding the relationship between valance and engagement. H1: A positive correlation should exist between message valance and political engagement. 16 Aside from message substance, characteristics related to the message source should further affect civic engagement. Specifically, people who are perceived to be credible political thinkers should be more persuasive in their recommendations to (dis)engage in political affairs. This relationship reflects Katz and Lazarsfeld’s two-step flow model, Robinson’s (1976) findings on political discussion partners, and Cohen’s (2003) experiments revealing the importance of source cues. In particular, independent of message quality, arguments coming from credible sources are more likely to be accepted than messages coming from non-credible sources (Chaiken, 1980; Petty, et al., 1981; Robinson, 1976). Source credibility is typically defined as perceptions of trustworthiness and expertise (Hovland, Janis, & Kelley, 1953). Trustworthiness refers to the belief that the speaker is being honest, unbiased, and fair. Expertise refers to judgments about the speaker’s knowledge in the subject matter such as experience, training, and qualifications. Taken together, this line of research suggests that credible sources are more persuasive. Thus, for this study not only is it important that the sources of memorable political messages are identified, but also that dimensions related to their political credibility are also ascertained. H2: Perceptions of source credibility and message valance will influence political engagement such that more positive message valance and higher source credibility will lead to higher levels of political engagement. In addition to credibility cues, another factor that is likely to be employed is perceptions of similarity. The importance of political similarity lies in the assumption that if people are similar, they will likely share common political self-interests. Experimental evidence in support of this assumption has revealed that African-Americans are more likely to vote for AfricanAmericans, women are more likely to vote for women, and young people are more likely to 17 endorse younger politicians (Sigelman & Sigelman, 1982). Further, research conducted by Bailenson, Garland, Iyengar, and Lee (2006), found that when candidate images were covertly distorted to look more like the participants involved in the research, participants rated that candidate more positively. Thus, similarity perceptions – even across arbitrary dimensions represent an influential political heuristic. Following this logic, if the source of participant’s memorable political message is rated high in similarity, it is more likely that participants would accept the message advocated by this source. This leads to the final hypothesis. H3: Source similarity, message valance, and source credibility will impact levels of political engagement such that as perceptions of similarity and source credibility increase, and as the message is rated more positive, levels of political engagement should increase. Method Overview There are two purposes for this research and the research design reflects these goals. The first aim is to create a working typology for the content of memorable political messages people glean from the media and discuss with one another. This typology will enable the testing of the research questions. The second purpose is to understand how these messages influence the subsequent political behavior and attitudes of the message recipients and tests hypotheses one through three. In order to achieve these objectives, data collection was carried out in two phases. In the first phase, approximately 85 memorable message surveys were collected. These messages were then used in the second phase as coding content for the card sorting procedure. Following the second phase, an additional 100 surveys were collected to increase statistical power. The methods and instrumentation used to achieve these objectives are described below. 18 Participants In the survey portion of this study 191 (n = 129 females) college students at a large Midwestern university were recruited for course credit in their Communication class. This sample had a mean age of 20.57 years old (SD = 1.66). Further, 75.5% of the sample identified as White, 7.4% Asian, 9.0% African American, 1.6% Latino, 1.0% of Mixed race, 5.6% of the sample identifying as ‘other’. Due to time and complexity considerations, for the card-sorting portion of the study, only the first 85 survey responses were included. During the card sorting portion of this study, an additional 86 college students (n = 66 females) were recruited from communication courses and similarly compensated for their participation. The mean age of this sample was 20.53 years old (SD = 1.45), with 85% identifying as White, 7% Asian, 4.7% African-American, and the remaining 3.4% identifying either as mixed race or ‘other’. Survey Procedure During the survey phase of this project, students were solicited in class and sent via email a link to an online survey. Participants were first asked to type their memorable political message. After their message was recorded, participants were then asked several questions regarding the context surrounding the message and source, along with their political engagement. The survey took students approximately 30 minutes to complete. Card Sorting Task Procedure In order to create a student-generated coding scheme, 86 participants were presented with a deck of filing cards, each with a different memorable message obtained from the survey portion of the study. Students were asked to sort the cards into separate groups on the basis of perceived similarity, and told that they could create as many groups as they like but each card could only 19 fall into one category (Kruskal & Wish, 1978; Rosenberg, Nelson, & Vivekananthan, 1968). On average, participants created 8.28 piles (SD = 2.66), with a range from 2 to 15 piles. This methodology allowed these groups to be analyzed using cluster analysis and multidimensional scaling by counting the number of times each message was placed in the same category. This technique has been established in past research to identify typologies in a way that is less susceptible to researcher bias (e.g., Deaux, Reid, Mizrahi, & Ethier, 1995). After participants sorted their cards into piles, they were asked to write a brief description of each pile’s similar characteristic. Independent Variables Memorable messages. Respondents were asked to complete a survey about memorable political messages. After reading through an introductory paragraph (Appendix A) participants were asked to “report a memorable message if one comes to mind concerning political affairs” using a script adapted from Smith and Ellis (2001). Out of the 191 participants completing the survey, two people (1%) reported not recalling a memorable message. Additionally, four participants’ messages were dropped because they were too difficult to decipher, resulting in a total of 185 valid messages. A scale was developed to assess ease of message recall. In order to assess message recall difficulty, four 7-point Likert scale items were summed and averaged (M = 5.28, SD = 1.16, α = .65), with higher scores representing greater ease with message recall. These items included: “My memorable message was easy to recall”, “The memorable message I recalled was the first to pop into my head”, “My memorable message was difficult to recall” (reverse scored), and “If I had to do this again, I would report the same memorable message”. Overall, participants rated their message as easy to recall, as evidenced by the sample mean falling significantly above the 20 midpoint (4) of this 7-point scale (t [182] = 14.71, p < .001). More specifically, only 8.2% of the sample scored a 4 or below on this scale, whereas 29.5% of the samples scored a 6 or higher on this ease of recall scale (11.5% scored a 7). These numbers provide self-reported evidence that most people not only have political memorable messages, but also that for a majority of people, these messages are readily accessible. Additionally, because the definition of a memorable message stipulates that the message be remembered for “long periods of time”, an open-ended question on the survey asked participants “when they heard the message for the first time?” Responses on this question indicate that, on average, the messages reported occurred 4.2 years ago (SD = 3.85 years). Further 22.5% of the sample reported a message from the past year, while only 1.6% of reported a message they heard within the last month. This finding does support the claim that the messages generated through this study are memorable in that the majority of recalled messages were recalled from at least one-year prior and with a range up to 19 years prior. Validity check. In addition to self-reported message importance, a follow-up survey was sent out to participants who completed the survey (n = 85) approximately 4 to 6 weeks after the original survey was distributed. The purpose of this follow-up survey was to test whether participants recalled the same message. If participants recalled the same memorable political message, evidence regarding the validity of these messages is supported. The first 85 participants completing the survey received an email solicitation to complete the follow-up survey between 4 to 6 weeks after completing the first survey. Of these 85 participants, 71 filled out this second survey constituting 84% of the eligible sample. Two questions were on the follow-up survey, the first asked participants, in identical language as the original survey, to recall their memorable political message and the source of this message. Both the original message and the follow-up 21 message were then compared using two trained coders. Coders scored a zero if the participant could not recall their message, recalled a different message, or recalled a message from a different source, a 1 if participants correctly stated the source of the message but did not go into further detail, a 2 if participants stated the correct source and paraphrased their message, and a 3 if participants recalled their message and source exactly. An inter-rater correlation revealed that 1 the two coders were highly reliable (r = .80) . Most participants were able to recall their memorable message almost perfectly. More specifically, only four people could not recall their memorable message at all (5.4%), and were given a zero by both coders. On the other hand 92% (n = 65) of the sample recorded a score of two or three by both coders indicating that they recalled the source of their message and, at a minimum, were able to paraphrase the same message they reported four to six weeks prior; thus supporting the resiliency of these messages. Memorable message content coding. In order to create the memorable message typology list, two strategies were used to code the memorable message data. The researcher derived one coding scheme, while the other was generated from students using a card sorting procedure (described below). The researcher derived coding scheme was created in a similar fashion to the card sorting procedure. Specifically, the researcher went through the messages serially, placing messages that were similar in the same category and different messages in an “other pile”. As more messages were read, the other pile was reduced until all messages fell into a created category. After this scheme was created using the first 85 messages, (see Appendix D) two trained coders coded all 191 messages according to the researcher’s scheme (Scott’s Pi = .72). This scheme included eight content categories and an “other” category. A card sorting procedure was the second procedure used to code the memorable messages obtained in the survey. The purpose of having research participants categorize these 22 messages was to examine whether the researcher generated scheme differed from student classifications in important ways. Of the 81 valid messages obtained through the survey, only unique messages were chosen for the card-sorting task. The reason for this was that substantively similar messages would be coded into the same category – so their use in the card sorting portion of the study would not provide any new information. From the set of 81 messages, 51 messages 2 were coded as unique and used in the card sorting task. All note-cards contained an identifying number and only one message. Additionally, all messages were edited to eliminate grammatical and factual errors, so that participants would focus on the message’s content and not the accuracy or message writer’s credibility when creating message categories. Participants were instructed to place cards into categories based on content similarity. They were further told that they could create as many categories (i.e., “piles”) as they would like, with the understanding that each card could only fall into one category. After participants sorted the notecards into piles, they recorded which cards fell into which pile, and described what characteristic, or set of characteristics, defined each category created. A dissimilarity matrix created from the card sorting task was arranged in a 51 X 51 matrix of nonoccurrence and submitted to ALSCAL (alternating least-squares scaling) using IBM-SPSS v.19. Multi-dimensional scaling (MDS) began after a Euclidean distance was specified in one-, two-, three-, four-, five-, and six-dimensional space (Kruskal & Wish, 1978). 2 The stress (and R ) values were: .53 (.24), .28 (.57), .17 (.76), .12(.85), .09 (.89), .09(.89), respectively. For this study, the four-dimensional solution was chosen based on interpretability 2 and diminishing returns in both stress and R values for the fifth and sixth dimensions (Kruskal 3 & Wish, 1978). The eleven categories that emerged through this analysis can be found in 23 Appendix D. After these categories were decided, two trained coders assigned scores to the remaining 106 messages not included in the card-sorting procedure, according to the scheme generated through multi-dimensional scaling (Scott’s Pi = .82). Message valance. Message valance was measured via self-report. Although this message characteristic could have been coded, in an effort to reduce error due to ambiguous data, valance reports were self-reported. Participants were first asked generally whether they considered their message to be positive (n = 115), negative (n = 30), or neutral (n = 46). Following this initial question, five semantic differential items gauged message valance. The anchors of these 7-point scales included: positive-negative, politically optimistic-politically pessimistic, favorable unfavorable, discouraging-encouraging, good-bad. Confirmatory Factor Analysis (CFA) using 2 Hamilton and Hunter (1988) revealed high internal consistency, χ (9) = 2.52, p = .980, e’s < .02, RMSE = .03, and strong reliability α = .94. Thus, this five-item scale was averaged and retained for analysis (M = 5.20, SD = 1.64), with higher scores indicative of more positive valance. A one-way ANOVA using the one-item measure for valance as the between-subjects factor, and participant’s score on the valance scale as the outcome measure demonstrated that participants who rated their message as positive also had the highest scores on the valance scale (M = 6.18, SD = 0.87), followed by the neutral messages (M =4.52, SD = 0.99), with negative messages receiving the lowest valance scores (indicative of most negative; M = 2.51, SD = 0.93), F (2, 185) = 207.06, p < .001. Message source. Four questions were created to assess whether messages were considered interpersonal or mass mediated in nature. These questions were created to gauge different dimensions relevant to interpersonal communication. These questions included: do you know the source of your message on a personal level (relationship dimension), have you 24 communicated with this person about other topics in the past (relationship dimension), was this message directed specifically towards you (message personalization dimension), would this person know you if you ran into them on the street (relationship dimension). If participants answered “yes” to all of these questions, then the message was considered interpersonal in nature (n = 27). If a participant answered “no” to any of these questions, then their message was classified as coming from the mass media (n = 163). In order to assess message channel, one multiple-choice question asked, “How did you receive this message?” with response options that included: face-to-face discussion (n = 49), phone (n = 1), email (n = 1), instant message device (n = 0), a text message (n = 0), TV (n = 101), radio (n = 0), website (n = 15), newspaper (n = 2), in-class (n = 12), at a rally or protest (n = 6), or on a sign, poster, or bumper sticker (n = 4). In order to assess the role of the source, a one item open-ended question asked, “Who was the source of this message the first time you heard it? A coding scheme was generated based on a random sample of responses and was accepted when responses no longer fell into the “other” category (Appendix D). Eleven different source roles were cited including: a politician, pundit, family member, friend or teacher, artist, activist, journalist, unknown person, propaganda piece, political advertisement, political materials, and an “other” category. A Scott’s Pi (.94) demonstrated high reliability between the two coders. Source credibility. Aside from obtaining the source of the memorable message, relational variables were also assessed including perceptions of credibility and similarity. Expertise and trustworthiness are often considered the two dimensions of source credibility (Hovland, et al., 1953). In order to construe measures of politically relevant expertise, three seven-point semantic differential items were asked including: experienced-inexperienced, informed-uninformed, and expert-non-expert (M = 5.10, SD = 1.24, α = .85), with higher scores 25 indicating higher expertise. Trustworthiness included a six-item semantic differential scale that included: biased-unbiased, honest-dishonest, trustworthy-untrustworthy, open-minded-closedminded, fair-unfair, and unselfish-selfish. A Confirmatory Factor Analysis using Hamilton and Hunter (1988) suggested that a four-item measure excluding the selfish item and honest item be 2 retained, χ (5) = 0.69, p = .984, e’s < .09, RMSEA = 0.13, α = .73. This four-item trustworthiness scale should be interpreted such that higher scores indicate higher levels of trustworthiness (M = 4.42, SD = 1.27). Because expertise and trust are typically considered the two dimensions of credibility, it was tested whether these two variables were second-order unidimensional. In order to do this, the correlation between these two scales was first assessed, r (187) = .53, p < .001. Due to this high correlation, these two scales were then tested against a third, unrelated scale. A CFA conducted using Hamilton and Hunter (1988) revealed that expertise and trust exhibited psychometric properties inherent in a second-order unidimensional factor model when one-item from the trust scale (bias-unbiased) was dropped. Based on these tests and an error assessment (e’s < .15, RMSE = 0.06), credibility was measured by calculating an average of six-items, three from the expertise, and three from the trust scale (M = 4.91, SD = 1.15, α = .82), with higher scores indicative of higher credibility. Perceptions of similarity. In order to measure similarity, participants responded on a 7point scale ranging from (1) very dissimilar to (7) very similar to the message source in the following areas: age, political knowledge, attitudes, education, personality, political affiliation, interpersonal style, political interests, religious beliefs, and income. These areas were adapted from O’Keefe’s (2002) review of similarity perceptions. After a test of internal consistency using Hamilton and Hunter’s (1988) CFA program, the full version of the similarity scale did not fit. 26 2 The scale chosen to represent participant similarity based on tests of internal consistency (χ (9) = 16.98, p = .05) and reliability (α = .75) was a five-item scale comprised of: level of education, personality, interpersonal style, political interests, and religious similarity. The scores were summed and averaged such that higher scores were indicative of higher similarity (M = 3.81, SD = 1.24). Outcome Variables Three scales were used measure political engagement, which were considered separately and indexed (Appendix B). These included a 13-item political participation scale (Verba et al., 1995), a five-item political knowledge test (Delli Carpini & Keeter, 1993; Shulman, Boster, & Carpenter, 2010), and a three-item general political interest scale. These scales were chosen to ascertain relevant democratic outcome variables. The well-established political participation index represents the behavioral component of engagement and was comprised of 13 activities in which participants stated whether they had (scored 1) or had not participated in that activity (scored 0). Examples of these activities include voting in elections, attending political rallies, and public displays of political beliefs (e.g., bumper sticker). Scores were summed and ranged from 0 to 13 (M = 4.08, SD = 2.80), and demonstrated adequate reliability, α = .73. A political knowledge test was used as a proxy to determine the quality of people’s political beliefs (see Zaller, 1992). Participants were asked five questions assessing their knowledge of political facts and figures, and were given a 1 for correct answers and a 0 for incorrect answers. Participants’ number of correct answers was summed to create a political knowledge test score (M = 2.02, SD = 1.40, α = .63). The general political interest scale was used to measure general interest in political affairs. Three questions measuring self-reported assessments of political knowledge, confidence 27 in one’s ability to recall political facts, and general political interest comprised this scale. These three questions were chosen due to their use in a previous study, which found that this scale was positively related to political knowledge and political participation (e.g., Shulman, Boster, Carpenter, & Shaw, 2011). Due to this scales convergent validity and strong reliability (α = .85), this scale was included in this study (M = 3.21, SD = 1.30). Additionally, because these variables were all significantly correlated with one another with one another (.19 < r < .22), an index consisting of the sum of the means from three scales was also calculated to create an engagement index, with scores ranging from 1.33 to 19.67, M = 9.32, SD = 3.92. Results Research Question One The first research question inquired what types of content would be recalled in people’s memorable messages. Content was classified based on a researcher generated coding scheme consisting of nine content categories and a student coding scheme generated through multidimensional scaling that consisted of ten content categories (the “other” category was removed from both schemes for the reporting of results). When comparing these two approaches, six content areas overlapped and several similar themes emerged. The general themes that emerged in both schemes were: media-related messaging, attitudinal or evaluative messaging, political beliefs, and one’s reflection on certain political events/experiences. The biggest difference between these two classification schemes was that the card sorting tended to prioritize context more than the researcher scheme. For example rather than creating a category to encompass all political campaign slogans, card sorting distinguished between President Obama’s slogans and slogans from all other politicians. Another example is that the card sorting tended to group all non-domestic affairs together, rather than distinguishing whether these affairs related to issues, 28 political parties, slogans, etc. Overall, however, the two methods produced more similarities than differences in message coding. These similarities are discussed below. Media related messaging. In this sample 35.8% of messages recalled came from the media, and more specifically, from political campaigns. Several students recalled President Obama’s slogan “Hope for Change” (n = 45), other students recalled a popular culture phrase (e.g., “live free or die trying”, n = 12), whereas some students recalled campaign slogans from their hometown or other local campaigns (e.g., “My man Mitch [Daniels]”, “One Tough Nerd [Rick Snyder]”, n = 11). For the researcher generated coding scheme, these media messages comprised two content categories. One category dedicated to campaign slogans (n = 56), and the other described as “non-candidate quotes” (n = 16). The distinction between these categories was that the campaign slogans were short, often repeated several times throughout a campaign, and were uttered directly by a politician. The quote category also consisted of short quotes, but the sender or creator of these quotes was not a political figure. This served as the key distinction between these categories. The card sorting coding scheme led to the creation of three categories that fell under this theme. Unlike the researcher generated scheme, students noted a distinction between Obama’s campaign slogans (n = 45), and slogans used by other candidates (n = 11). Thus, the primary difference between these messages was whether or not the author was Obama– yielding two slogan categories. Card sorting corroborated the “non-candidate quote” category (n = 12), which placed quotations not from a politician in the same category. Attitudinal messaging. Message categories that fell within this theme incorporated evaluative statements, and consisted of approximately 38% of the sample. Specifically, these messages contained information about whether a political party, politician, or political issue was 29 good or bad. Within the researcher classification scheme four content categories could be considered evaluative in nature. The first content category consisted of students who recalled the negative aspects of politics. This includes recalling campaign attack ads or fear appeals (e.g., one student recalled the GOP questioning John Kerry’s swift boat story), or alluding to a message about “evil” or “out of touch” politicians (n = 24). Students expressing frustration with the current political system defined the second content category. These messages contained expressions about the lack of solidarity within the government and bemoaned the growing divide in American politics (n = 5). Contrary to these pessimistic political messages, pro-democracy sentiments were also recalled by students (n = 29), and comprised the third category. These messages included quotes and conversations that centered around the importance of voting and general engagement, and also included proud moments in political history such as John F. Kennedy’s infamous quote from his inaugural address, and Martin Luther King Jr.’s “I have a dream” speech. The fourth researcher generated category included in this theme were messages in which students thought through a political issue and why one side was right and/or the other was wrong. Some of these issues included: food chain management, the war in Iraq, gay marriage, and Japanese whaling (n = 15). Within the card sorting scheme, the negativity surrounding politics (n = 25), prodemocracy sentiments (n =16), and political issue debate categories (n = 13) also emerged. These were identical to the researcher coding scheme described above. Interestingly, however, the researcher and card sorting scheme diverged in one respect. In the researcher scheme, students who bemoaned the divisive nature of politics were placed in one category. In the card sorting scheme however, a category emerged that emphasized the need for unity rather than the existence of division. Students placed the divisive messages, described above, in the negative 30 political messages category. They created a different category for messages that called for unity and expressed the need for America to come together (n = 18). Examples of messages placed in this category were the Martin Luther King Jr.’s “I have a dream speech”, and Abraham Lincoln’s quote “A house divided against itself cannot stand”. Political belief messages. Political belief memorable messages include content about what a person believes to be true or false (6.3% of the researcher scheme and 11.6% of the student scheme). In both the researcher and student generated coding scheme, beliefs about political parties emerged as a category (n = 12 & n = 15, respectively). Specifically, through these messages, students recalled beliefs about what attitudes or beliefs defined the political parties. For example, one student stated that the difference between democrats and republicans was that “democrats are liberal and open-minded, but might not be that wealthy. Republicans are wealthier and more narrow minded and conservative with social issues”. Thus, messages in this category include descriptors about what defines a political person or issue. For the student generated coding scheme, one additional category fell within this theme. Students coded all nondomestic issues as being similar to one another (n = 7). Included in this category are beliefs about: whaling in Japan, the situation in Egypt, and Hu Jintao’s visit to the US. Political Experiences. The final theme emerging in both coding schemes included personal recollections about a message that occurred during a political event (14.7% of sample). In the student coding scheme, students distinguished between messages referring to September th 11 (n = 8) and all other political events (n = 16). In the researcher coding scheme, this category included all messages relating to a political event or experience such as attendance at a rally or th September 11 imagery (n = 28). 31 It should be noted that only two students stated that they had no political message. Additionally four student’s messages could not fit reliably within either coding scheme and were therefore regarded as “other” in both coding schemes. Research Question Two The second research question inquired whether participants would be more likely to recall a positive, negative, or neutral political message. In order to address this question, participants’ response to the one item general valance question was tested. In all, 30 participants recalled a negative message, 46 neutral, and 115 were positive messages. A Chi-Square test found that the frequency difference in participants reporting positive, negative, versus neutral 2 messages was statistically significant, χ (2, n = 191) = 64.10, p < .001, demonstrating that the odds of recalling a positive memorable political message are approximately 2.3 times greater than recalling a neutral or negative message. Research Question Three The purpose of research question three was to examine the source of people’s memorable political messages. This was done in two ways. The first way assessed whether, in general, memorable messages came from interpersonal or mass media sources. Results show, that participants, overall, were significantly more likely to report a message coming from the mass 2 media (n = 163) than from an interpersonal source (n = 27), χ (1, n = 190) = 97.34, p < .001. Put differently, the odds of a person reporting a message from the mass media were 35.5 times greater than the odds of participants reporting a message from an interpersonal source. In addition to investigating whether a person recalled a mass media or an interpersonally transmitted message, the role of each message source was also examined. By far, students were most likely to recall a message from a politician or an ex-politician (23.6% of all message 32 sources). Interestingly, the next two sources most frequently cited were messages from family members (16.2% of all messages) and from friends (13.1% of messages). It bears mentioning that 26.2% of the sample’s sources fell into an “other” category. This category was created because many students were vague and cited “the TV” or “a commercial” as their source. The remaining 23.5% of the data were roughly equal in proportion and consisted of sources that included: pundits or journalists (6.3%), artists or celebrities (2.1%), activists (1%), political propaganda such as posters or t-shirts (2.6%), political materials such as books, films, and documentaries (3.7%), and political campaign ads with an unspecified source (3.7%). Research Question Four The purpose of research question four was to test whether certain sources were more or less responsible for certain types of messages. This research question was tested using an 2 omnibus chi-square test, χ (110, n = 190) = 254.15, p < .001. In order to look at this more closely, the three largest source categories (politician or ex-politician, parent, and friend or teacher) were compared against one another (n = 101). All other sources were excluded due to small n’s (n < 12) or lack of specificity (in the case of the “other” category). Again, the chi2 square test was significant, χ (16, n = 101) = 64.23, p < .001, indicating that certain sources are more likely to send certain types of messages. The pattern of data reveals that, not surprisingly, politicians are more likely to produce a campaign slogan (n =26) than family (n = 0), friends or teachers (n = 3). On the other hand, messages that create a political archetype commonly come from the family (n = 9), rather than politicians (n = 0), or friends or teachers (n = 1). The most common message coming from friends and teachers were pro-democracy messages (n = 10), followed by politicians (n = 5) and family members (n = 4). Finally, it appears that memorable negative political messaging is more likely to come from family members (n = 5) and friends (n 33 = 4), than politicians (n = 0). Taken together, these findings suggest that certain messages can be attributed to specific sources. The pattern of results described above leads to interesting implications regarding political socialization messages (in the case of political archetypes spread by parents), the resolve of education to facilitate civic engagement (in the case of pro-democracy messaging), and the ability of campaign designers to create memorable candidate slogans. Research Question Five Once content categories were created, it became important to examine whether a person’s memorable political message associates with their civic engagement. In order to test this question, four one-way Analysis of Variance (ANOVA) tests were conducted using message category as the between subjects factor, and each of the four engagement variables (knowledge, participation, interest, and engagement index) as the dependent variable. None of these analyses were statistically significant, indicating that the type of message recalled by students does not clearly relate to their level of civic engagement. One explanation for these null effects is that several of the content categories had small n’s (n < 10), greatly hindering an ANOVA’s ability to test significant differences due to low statistical power. The means generated by condition are presented in Table 2. Hypothesis One Hypothesis one predicted that a significant correlation would exist between message valance and political engagement, such that people recalling more positive messages would be more likely to be civically engaged. In order to test this hypothesis, four correlation analyses were conducted testing the relationship between message valance and political knowledge (r (187) = -.03, p = .651), political participation (r (186) = .02, p = .845), general political interest (r (187) = -.09, p = .234), and the engagement index (r (186) = -.03, p = .737). Taken together, 34 these results suggest that the valance of people’s memorable political message does not relate to people’s actual political engagement. Thus, hypothesis one was not supported. Hypothesis Two Hypothesis two stated that source credibility and message valance would influence civic engagement. Specifically, as participants report more positive messages coming from highly credible sources, political engagement should increase. This hypothesis was tested using multiple-regression with the continuous measure of message valance, source credibility, and a cross-product interaction term between these two predictors, along with the four civic engagement dependent variables. Thus, four models were ran and can be found in Table 3. For the models predicting political knowledge and political participation, the omnibus models, individual predictors and interaction terms were not significant, thus failing to provide support for hypothesis two. The model predicting political interest, however was statistically significant, F (3, 181) = 4.03, p < .01; furthermore, message valance (b = -.82, se = .20, t = -3.20, p < .01) and the interaction term (b = 1.01, se = .04, t = 2.65, p < .01) were also significant. Specifically, and despite predictions, as memorable messages became more positive, level of political interest decreased. Similarly, although the omnibus model predicting scores on the engagement index was not significant (F (3, 182) = 1.88, p = .135), message valance (b = -.61, SE = .63, t = -2.35, p < .05), credibility (b = -.36, SE = .71, t = 1.78, p = .08), and the interaction term (b = .88, SE = .13, t = 2.27, p < .05) were all statistically significant but not in the predicted direction. Overall, these findings fail to provide support for hypothesis two. Two out of the four models found no relationship between message valance and source credibility on political engagement. In the models that did produce significant effects, results were in the opposite direction, which suggests 35 that as memorable messages become increasingly negative, and as sender credibility decreased, political engagement scores increased. These findings are displayed in Table 3. In order to examine the significant interaction effects more closely, means on the political engagement scales were assessed under varying levels of message valance and source credibility. The valance variable tested here was the one item measure that asked participants to state whether their message was positive, negative, or neutral. Source credibility scores were trichotomized based on means, to indicate whether participants judged their message source as high, medium, or low in credibility. Using these variables, a 3 (valance: positive, negative, neutral) x 3 (credibility: high, medium, low) Analysis of Variance (ANOVA) was run. When political interest was the dependent variable, the omnibus ANOVA was not significant, F (8, 2 172) = 1.15, p = .331, Adj.R = .01, nor were any of the main or interaction effects. This pattern of results can be found in Figure 1. When the dependent variable was the engagement index, the omnibus model was also not significant, F (8, 171) = 1.33, p = .231. The interaction term, however, was statistically significant, F (4, 171) = 2.58, p < .05. Contrary to expectations, political engagement levels were highest when participants recalled a negative message from a low credibility source (M = 10.92, SD = 0.99), and engagement levels were lowest when participants recalled a negative message from a high credibility source (M = 6.78, SD = 2.28). These findings are depicted in Figure 2. Finally, another test looking at the effect of message valance and source credibility examined the strength of the correlation between message valance and engagement under low, medium, and high conditions of source credibility. These correlations can be found in Table 4. Interestingly there was a significant negative correlation between message valance and political interest and the engagement index under conditions of low source credibility. Specifically as 36 messages become more negative, interest and general engagement increases. This pattern is in the opposite direction however, when a source is considered high in credibility – such that the more positive a message becomes, the more engaged the person claims to be. In sum, these results suggest that political interest and the engagement index are influenced by message valance and credibility. It appears that when a person seemingly disagrees with their message (defined by a person recalling a message that they consider negative and from a lowly credible source), engagement scores tend to be higher; moreover, when a person recalls a positive message from a highly credible source, their engagement levels tended to be higher as well. Hypothesis Three Hypothesis three predicted that in addition to source credibility, and message valance, perceptions of similarity with the message source should also affect participants’ political engagement. In particular, it was thought that when a participant’s memorable message is positive and from a highly credible and similar source, political engagement should be highest. Similar to analyses in hypothesis two, these relationships were tested in several different ways in an effort to thoroughly explore patterns. Thus, the first analysis conducted was multiple regression, followed by an ANOVA, and finally a correlation analysis. Four multiple regression models were ran to assess whether valance, credibility, and similarity predicted level of political engagement. For all models these three predictors were entered into the first block of the model. In the second block, the four cross-product interaction terms between these variables were entered. All models can be found in Table 5. The omnibus 2 model predicting political knowledge, F (7, 175) = 1.07, p = .385, Adj. R = .00, and political 2 participation, (7, 174) = 1.22, p = .294, Adj. R = .01, were not significant. The models 2 predicting political interest, F (7, 175) = 3.99, p < .001, Adj. R = .10, (similarity, b = .69, SE = 37 2 .34, t = 2.16, p < .05) and scores on the engagement index, F (7, 174) = 2.45, p < .05, Adj. R = .05 were statistically significant. These findings suggest that when predicting political interest, sources that are rated higher in similarity correspond with higher levels of engagement. Aside from this finding, however, overall these findings proffer little support for hypothesis three. In order to examine specific trends more closely for the models predicting political interest and overall engagement, a between-subjects independent groups 3 (valance: positive, negative, neutral) X 3 (credibility: high, medium, low) X 2 (similarity: low, medium, high) ANOVA was ran. The omnibus model means on the political interest scale nearly achieved conventional levels of statistical significance, F (25, 155) = 1.55, p = .056. Within this model, a main effect for source similarity emerged, F (2, 155) = 3.14, p < .05, such that political interest levels were highest when participants recalled a message from a highly similar source (M = 3.69, SD = .25), followed by a source low in similarity (M = 2.92, SD = .21), and moderately similar source (M = 2.86, SD = .23). Additionally, although failing to reach conventional levels of statistical significance, there was a modest credibility by similarity interaction effect, F (4, 155) = 2.02, p = .09, such that interest levels were highest when participants recalled a message from a highly credible and similar source (M = 3.62, SD = .39), and interest levels were lowest when participants recalled a message from a source low in credibility but moderate in similarity (M = 2.59, SD = .30). Although not predicted, these patterns are similar to those revealed in hypothesis two. Namely, that people who are most engaged tend to recall either a) highly positive messages, from credible and similar sources (the predicted relationship) or b) tend to recall negative messages that come from lowly credible and dissimilar sources. The omnibus model examining engagement scale means also did not quite reach 2 conventional levels of significance, F (25, 162) = 1.55, p = .056, Adj.R = .07. The only factor to 38 approach statistical significance was an interaction effect between source credibility and message valance, F (4, 154) = 2.22, p = .07, such that participants who recalled a negative message from a source low in credibility registered the highest total engagement (M = 10.93, SE = 1.44), followed by those who recalled a neutral message from a moderately credible source (M = 10.26, SE = 0.89). Conversely, participants who recalled a negative message from a highly credible source had the lowest engagement scores (M = 6.75, SE = 2.35). Although these patterns deviate from initial predictions, these findings do suggest that message valance and message source interact in complicated ways to affect participant’s political engagement. In addition to these models, Table 6 displays the strength of the correlation between message valance and political engagement variables under different levels of source credibility and similarity. Interestingly, germane to this investigation, when source credibility is perceived as low, the relationship between message valance and political engagement is in the opposite direction predicted by all hypotheses (.04 < r < -.73). Specifically, these coefficients are all negative suggesting that as the message becomes more negative levels of engagement increase. On the other hand, those who recalled messages from a highly credible source exhibited patterns of relationships consistent with hypotheses that proposed that people recalling more positive messages would also be more likely to be politically engaged (-.12 < r < .42). Furthermore, when subjects reported a message from someone who was perceived as very dissimilar and low in credibility, the negative relationship was augmented (average r = -.30); additionally, when participants recalled a message from someone who was very similar and highly credible, the predicted positive relationship was apparent (average r = .20). Taken together, some of these findings are consistent with hypothesis three while some are not. Although the relationship between highly credible and similar source messages led to the predicted positive correlation 39 between engagement and message valance – what was unexpected was the population of people who recalled a negative message from a lowly credible (and dissimilar) source was just as, if not more, engaged. These findings will be discussed in greater detail in the subsequent section. Discussion The purpose of this study was to investigate memorable political messages and to test if certain messages are related to political engagement. The contributions made by this research were to first, develop a classification scheme for different types of political messages, and second, to foster a deeper understanding of the relationship between message substance and political outcomes of interest. Guided by Burleson’s (2009) message centered approach to interpersonal communication, the substance, and manner with which memorable political messages were sent were expected to relate to a person’s political engagement. By obtaining participants’ memorable political message, coding the substance of these messages, and considering the credibility and similarity of the message source, it was hoped that a greater understanding would be gained for why people become politically involved. The first priority of this paper was to create typologies of memorable political messages. The reason for this pursuit was because in the political communication literature, few studies investigate naturally occurring communication content. Thus, first examining what people think about when they think about politics represents an extension of a large line of research. In order to generate this information, participants were asked to, “recall and record what message comes to mind when you think about political affairs?” Of the 191 participants, only two people stated that they did not have a memorable political message (1%). This non-response rate of one percent is smaller than the rates obtained in other memorable message domains [e.g. Smith & Ellis, 2001 (7% non-response); Smith et al., 2009 (40% non-response rate)], suggesting that 40 politics is a fruitful avenue to test these messages. Furthermore, in a follow-up survey four to six weeks later, participants were again asked to record their memorable political message. In this follow-up 92% of those responding reported the same message as weeks prior. This validates the significance, or at least memorability, of the messages collected in this study and warrants their current use in the creation of a content classification scheme. Discussion of Results Four content themes emerged from the memorable messages obtained regardless of whether coding was based on a researcher-generated scheme or through multi-dimensional scaling. People either reported a message from the media, and more specifically from a political campaign, a statement about their generalized attitudes towards political affairs, an expression of their political beliefs, or a recount of their political experiences. Not surprisingly, campaign messages from President Barack Obama’s political campaign was the most frequently reported message type. The reason this result is not surprising is due to the recency of the campaign (< 2 years ago), the repetition of these slogans, and because these slogans were relatively short and consequently easy to remember (Knapp, et al., 1981). Although different message content categories emerged, there was no pattern of association between message type and engagement means (research question five). One possible reason for the null effects, however, was due to low statistical power. Given that the sample consisted of 191 participants, and that 10 message categories were created, cell sizes consisted of 19 participants on average (6 < n < 45). Low cell sizes in combination with unequal numbers in each cell hindered the ability of statistical tests to detect significant results. Thus, although preliminary research suggests that message content does not correspond with engagement – small cell sizes obfuscate the validity of this conclusion. In the future, more data and more messages should be collected. 41 The results from research question two indicated that participants were twice as likely to recall a positive political message rather than a negative or neutral message. A post-hoc explanation for this positivity effect, however, is that participants rated campaign slogans positively – because usually slogans are crafted to be optimistic and positive. This would suggest that the presence of the campaign slogan message category is responsible for the positivity effect. To examine whether this was the case, this category was removed and valance was reassessed. Again, results suggest that positive messages are more likely to be recalled (n = 68) 2 than negative (n = 26) or neutral (n = 35) messages, χ (2, n = 129) = 22.71, p < .01. Thus, this effect was not isolated to campaign slogans, and instead represents a general trend towards recalling positive political messages. This finding replicates the original work on memorable messages that similarly found participants were more likely to recall positive, or benevolently stated, messages (Knapp et al., 1981). A positive implication is that this finding signifies a general sense of optimism among students regarding political affairs. Whether this effect is limited to this young demographic becomes an interesting empirical question. In addition to message substance, the manner in which these messages were delivered was also considered. Questions were created to assess the message source, the relationship between the source and receiver, and the channel through which the message was communicated. To identify the message source, participants were presented with an open-ended question, asking, “Who was the source of this message, the first time you heard it?” Approximately 45 students stated that they heard the message on TV, compared to 30 who said they heard their message from a parent, and 25 from a friend. This suggests that 45 people recalled a mass media message compared to 55 who reported an interpersonally transmitted message. These frequencies, however, differ unexpectedly from the multiple-choice question regarding message 42 channel. For this question, 101 students reported hearing their message initially on TV, whereas 51 stated they heard their message either face-to-face, or via phone or email. The remaining 39 participants reported hearing their message either online, during a rally/protest, in class, or in the newspaper. In an effort to clarify these discrepant results, four questions were asked about participants’ relationship with the source of their message. Based on results from this interpersonal relationship scale, only 27 people received an interpersonal political message compared to 163 students who reported hearing their message from someone they did not know very well, who also might not have been specifically talking to them. Thus, these measures lead to equivocal conclusions regarding whether certain message sources should be deemed interpersonal or mass mediated. For instance, although 55 participants reported hearing their message from a friend or family member, findings from the interpersonal relationship scale suggest that only 27 participants received their message personally from a known source. This measurement problem exemplifies a concern in the interpersonal communication literature regarding how interpersonal communication should be defined and measured. For example, if a person’s friend iterates Obama’s position on key issues, is the true message source the friend or Obama? Students struggling with identifying the initial source of their memorable message might explain discrepancies surrounding source, channel, and relationship responses. On the four item scale measuring the relationship between the message source and receiver the questions included: did you know the source of this message on a personal level, have you communicated with this person about other topics in the past, was this message directed specifically towards you, and would this person know you if you ran into them on the street? The purpose of this scale was to measure requirements from Burleson’s (2009) definition of interpersonal communication, namely that two people, with an established relationship, were 43 communicating directly with one another, with the possibility for feedback and shared meaning. Scores on this index (score 0 for no, and 1 for yes) ranged from 0 (n = 62) to 4 (n = 25), with 42 participants falling between 1 and 3. A closer look at the frequency distributions for these items does not clarify this discrepancy. For example, only 43 people said that they have communicated with this person in the past, whereas 54 people stated that the message source would know them if they ran into them on the street. This leaves an odd 11 person differential between people who know one another yet have never had a conversation before – and represents a difficult measurement problem to explain. One way to remedy this concern is by treating interpersonal communication as a continuous rather than dichotomous variable (Miller & Steinberg, 1975). This theoretical and methodological consideration will be returned to in the subsequent section. The relationship between message content and message source was explored in research question four. Rather unsurprisingly, political slogans were more frequent from politicians than family, friends, and teachers. The patterns that emerged in other domains, however, were more interesting. The family tended to be responsible for crafting archetypal images of the political parties. Political socialization research (e.g., Beck & Jennings, 1991; Jennings & Niemi, 1981) has found that parents are often accredited for transmitting political attitudes to their children. Findings from this study suggest that parents might effectively influence children’s political thought by creating and fostering party stereotypes. Students reporting a message that fell into this content category recalled messages about what it means to be a democrat or republican. Examining these messages further demonstrated that party descriptors tended to be overly simplistic and argumentatively weak in nature (e.g., “Republicans are rich, whereas democrats stand for the working class”). Perpetuating these simple stereotypes should aid the parent in 44 effectively transferring their political attitudes to their child, providing that the child is less exposed to alternative perspectives (Stoker, 2007). Another noteworthy trend was that negative political messages tended to come from either the family, or from the “other” source category. The idea that families perpetuate negative political attitudes is counterintuitive. Initially, it was expected that negative attitudes would come from political attack ads – which are often criticized for “turning off” or disenchanting constituents (Ansolabahere & Iyengar, 1995), or from political pundits often criticized for polarizing the electorate. Similarly, the presence of political scandals also should motivate people to think negatively about politics. Of the 30 messages identified as negative only 5 came from a politician or political advertisement, whereas 8 came from the family, and 3 from a friend or teacher. The remaining 14 messages came from an “other” source (a category created because people were not specific about their source, e.g., “TV”), activists, or political materials. The frequencies presented here are small so claims about their generalizability are not possible. If more data were collected, it would be interesting to test whether these trends would uphold. For all hypothesis testing, it was examined whether message substance, in terms of message valance, and manner in terms of source characteristics (e.g., credibility and similarity) associate with levels of political engagement. Political engagement was measured in four ways: scores on a political knowledge test, an index of political participation activities, a self-reported political interest scale, and the sum of these three scales comprising a comprehensive engagement index. For all hypotheses, a positive relationship was predicted between message valance and political engagement. It was thought that when a person recalled a message that they considered politically negative, they would also be less likely to engage in the political process. Conversely, people recalling politically positive messages should be more likely to be politically 45 involved. Despite this prediction, hypothesis one did not receive support for any of the political outcome measures. Results from subsequent hypotheses, however, revealed that the relationship between valance and engagement is far more nuanced. For example, hypothesis two stated that source credibility along with message valance would associate with political engagement. For the political interest dependent variable, and the comprehensive engagement index, the results were significant but in the direction opposite of predictions. On the one hand, it was argued that high credibility sources would lead to higher levels of engagement; furthermore, a high credibility source offering a politically positive message should be recalled by participants scoring highest on the engagement scales. Despite these predictions, results indicated that participants who recalled a negative message, from a lowly credible source were just as, if not more politically engaged and interested, compared to those recalling a positive message from a credible source. Further, when considering message source, a negative relationship emerged between valance and engagement, such that as the message a person reported was considered more negative, and as credibility was perceived as lower, the higher that person’s political interest and overall engagement. Although not predicted, this relationship is logical. Perhaps what drives people’s political interest, and motivates them to politically participate is not their acceptance and support of current political happenings, but rather their disappointment or disagreement with certain policies or politicians. This assertion runs contrary to a substantial line of public opinion literature that claims creating a negative political environment, or a highly partisan political environment, breeds disengagement and disenchantment in the political process (Ansolabahere & Iyengar, 1995; Mutz, 2002, 2005; Noelle-Neuman, 1974). 46 The final hypothesis included perceptions of source similarity, along with message valance and source credibility, to better understand how messages relate to engagement. Similar to previous findings, source credibility and similarity mattered, but not in obvious ways. Although main effects and interaction effects were generally unsupported, perhaps due to low statistical power – when looking at the correlations between message valance and engagement under different levels of similarity and credibility, interesting trends emerge. When participant’s memorable political message came from a source perceived as low credibility, there was generally a negative relationship between valance and engagement. Specifically, negative messages were more strongly associated with higher levels of interest and engagement. This negative relationship was especially pronounced when the low credibility source was rated high in similarity. Alternatively, when a source was rated high in credibility, there was generally a positive relationship between message valance and engagement – the association originally predicted. This positive correlation was most pronounced when the source was high in similarity, and when political knowledge was the outcome variable. Overall, these findings, though not predicted, illuminate noteworthy relationships between memorable political messages and political engagement. Although at face value, valance and engagement were not associated, results obtained suggest that the relationship between these constructs is much more nuanced. Methodological and Theoretical Insight This project intended to fill a gap in the political communication literature, namely that studies rarely examine what messages people think about when they think about politics. The majority of students recalled a political campaign slogan – however, and perhaps more interestingly, many others recalled messages that spoke to more personal issue positions or beliefs about the political parties or politicians. Because this study obtained actual and influential 47 messages, rare insight is afforded into why people choose to engage or disengage in political matters. The methodology used to attain the relationships between message, source, and outcomes was markedly different from other political communication research. Specifically because, unlike previous studies, self-report assessments of frequency of political talk or message stimulus materials did not comprise the data. Rather, participants were asked to think of a message that represented how they conceptualized politics. Interestingly, there were significant associations between the messages recalled and the outcomes reported in this sample. An important theoretical contribution is the extension of the memorable message paradigm to political affairs. Although with this extension some concerns about the memorable message literature became apparent. Memorable message scholars argue that people hold certain messages that are a) memorable for long periods of time, and b) influence beliefs, attitudes, and or behaviors in a particular domain. This study revealed that 98% of participants reported having a memorable message, and of those sampled, 92% of students were able to recall this message four to six weeks later. Additionally, participants tended to recall older, rather than more current, messages. On average, participants recalled a message that was first uttered 4.16 years ago with 15% of the sample reporting a message that first occurred over 10 years ago. These values when contrasted with only the 22.5% who reported a message occurring within the last year suggest that the messages obtained in this study are consistent with the time component of the conceptual definition of memorable messages. Despite these achievements and cursory support for applying memorable messages to the political domain, some paradigmatic concerns remain regarding the existence of memorable messages that merit theoretical consideration. The first concern pertains to the problem of defining a construct by its outcomes. O’Keefe (2003) points out the tautology of defining message characteristics (e.g. a fear message) 48 based on whether the message produces the intended effects (e.g., fear is induced). The problem with this thinking is the inability for the null hypothesis to be affirmed, because if the predicted outcome is not observed, then the message content cannot be (by definition) what the researcher set about to create. Thus, referring back to the example of a fear appeal, an ineffective fear appeal cannot exist by virtue of its definition that fear messages elicit fear. Relatedly, the definition of a memorable message suffers from the same problem. A message that is remembered for a long period of time, but does not influence beliefs, attitudes, or behaviors, cannot be a memorable message by definition. This leads to a problem of definition for those interested in this line of research to carefully consider, because certainly not all remembered messages impact behavior in ways that are easily testable. Without the second component of the memorable message definition, what is left to explicate regards a) what makes a message memorable versus not memorable, and b) what constitutes a message? In order to address the first question, guidance from the literature on memory was consulted. According to Schacter (1999) there are seven reasons why memory sometimes fails. These seven processes fall broadly into three categories including: forgetting, misremembering (or distortions), and intrusive recollections. For the purposes of this project, messages that are not memorable fall into the “forgotten” category and therefore the reasons why people forget will be the focus of this discussion. When thinking about characteristics of non-memorable messages, Schacter’s (1999) idea of transience is most applicable. Transience refers to when, “memory for facts and events become less accessible over-time” a process akin to gradual forgetting which is strongly associated with time. The inaccessible information may either be difficult to retrieve due to interference (i.e., “blocking”), or can actually be erased from memory due to lack of use. Given 49 the fact that the likelihood for memory retention decreases over time, what is interesting about the findings from this study is that only 22.5% of the sample reported messages occurring within the last year. This percentage compared with the 56.5% who reported a message uttered three or more years ago, and the 15% reporting a message from over 10 years ago supports the resiliency and somewhat non-random nature of these messages over time given the fact that, according to Schacter, information that is unused over long periods of time is more likely to be forgotten presumably due to the loss of synaptic connectivity (Bailey & Chen, 1989). Another reason why a message may not be memorable is because at the time of the message, participants were not devoting sufficient attention to the information (i.e., absent mindedness, Schacter, 1999). Applying this logic to the current study, with the 24-hour news cycle and campaigns that last over two years, some political messages may simply be ignored because people are reluctant and unable to constantly devote cognitive resources to public affairs (Downs, 1957). This process explains why the messages participants tended to recall included th widely publicized and important political events (e.g., September 11 attacks, the Presidential Inauguration) or were messages that were repeated and thus made salient (e.g., a political issue that was discussed on TV and then repeated by friends or parents). Based on this information, the reason why some messages are memorable and some are not is not straightforward. What can be deduced, however, is that messages that were uttered a long time ago are less likely to be remembered than recent messages. A caveat to this claim, however, is that a message is more likely to be remembered to the degree that it is used. Thus, messages that are often made salient due to context or repetition are more memorable than messages only stated once. Additionally, Schacter (1999) points out that features irrelevant of message content also aid or inhibit the likelihood of retention. Specifically, if a message has to 50 compete in a flooded message environment, or if the message occurred when someone was distracted it is unlikely that the message will be remembered for an enduring period of time. The second issue related to defining memorable messages was considering what constitutes a message. In this study, some participants did not recall what traditionally would be thought of as a message and instead referenced a political experience (e.g., attending an Obama rally). This begs the question of whether these responses validly represent a memorable message. Technically, a message is a symbolic exchange of meaning between people through a definable channel. Given no source explicitly exists when a person recounts an experience, the conceptual issue of whether this constitutes a memorable message merits debate. On the one hand, a person’s experience has the tremendous potential to impact and explain one’s beliefs, attitudes, and behaviors. On the other hand, because experiences are completely internalized, these responses could fall outside of the communication domain and instead could be more explainable using theory from cognitive psychology. Despite this issue of definition, and what to consider a valid versus invalid memorable message, what is important to note is that political experiences impact people undoubtedly because they are highly personalized in nature. Although experiences cannot be used as stimulus materials in communication studies, encouraging people to discuss their political experiences may be a way to engage audiences and personalize the political process. This latter solution clearly falls within the communication domain regardless of whether the data that originally informed this idea is considered communicative in nature. Aside from the aforementioned theoretical concerns with memorable messages, an interesting contribution was the finding that negative political messages were as, if not more, strongly associated with political interest and general engagement than positive message. The relationship between message valance and outcomes has been a source of controversy in the 51 literature. Ansolabahere and Iyengar (1995) for example used experimental evidence to argue that attack ads, or “negative campaigning”, disenfranchise populations by turning them off of the political process. Further Mutz (2002, 2006) claims that being in an environment that consists of heterogeneous political opinions inevitably decreases the political participation of audiences exposed to multiple perspectives (Participatory Democratic Theory). Despite these claims, however, a meta-analysis conducted by Lau et al. (1999) found no clear evidence suggesting that negative messaging leads to negative outcomes. Thus the authors concluded that, despite the prevalence of political attack ads, no evidence existed to suggest that positive or negative messaging was more effective at eliciting participation. Contrary to aforementioned theory and evidence, the current research found that participants who recalled a message that they considered to be negative exhibited higher levels of political interest and engagement. This presents a theoretical departure from the norm, and leads to speculation regarding why this might be the case. The first reason could be attributable to different methodology. In Mutz’s line of research (2002, 2006), for example, information is collected about a person’s immediate social network (e.g., asks a person to name up to five people with whom they talk about politics), and then a network discrepancy index is created to represent the degree of opinion heterogeneity in one’s political environment. Her primary dependent variable is participants’ ability to create arguments on both sides of a political issue. Mutz finds that people with more heterogeneous networks are more balanced in their knowledge of issue positions (2002). However, a troublesome fact for democracy is the additional finding that people who are exposed to differences of opinion (presumably a good thing for a democracy) are also less likely to participate in public affairs. 52 Alternatively, in the current study, participants were not asked to think about the breadth of political opinions, or messages, in their environment, but rather to espouse one representative political message. Regardless of one’s message environment, this research shows that when participants think of one, negative message from a non-credible and dissimilar source, they also tend to rate high in political interest and general engagement. This is not necessarily different than Mutz’s work, but does present an alternative way to think about the process of engagement. Namely, that it is not the mere presence of contrary opinions, but rather a particular contrary opinion that may motivate engagement and spark political interest. Another notable difference between this study and the research by Ansolabahere and Iyengar (1995) and Lau et al. (1999) was in the conceptualization of negative and positive messaging. In this study, a negative message was self-reported as negative by the message recipient. In most cases, the participant reported the message as negative because either they disagreed with the message (similar to Mutz, 2002, 2006), or because the message contained a negative evaluative statement (“politicians are bad”). This differs slightly from Ansolabhere and Iyengar’s (1995) and Lau et al. (1999) who specifically defined negative messages as attack ads used within the context of a political campaign. Within the context of a political campaign, Westen (2007) argues that attack ads are relatively ineffective because when participants view an attack ad about their preferred politician, they become emotional and immediately refute claims regardless of the evidence presented. Contrarily, when viewing a negative ad about an opposing side’s politician, people become systematic in their reasoning and allow that advertisement to reinforce their already negative opinion. Thus, in both cases, no opinion or participation change is likely to occur. In this study, however, when negative messaging was removed from a campaign context different results became apparent. Perhaps outside of a campaign context, 53 where attack ads are expected and source cues are particularly salient, audiences are more heavily influenced by the content of these messages. This assertion, however, merits further investigation. Another theoretical contribution was utility offered by Burleson’s (2009) message centered approach to interpersonal communication. Although interpersonal communication was not the primary channel through which messages were transmitted, importantly, his components of manner and substance did influence outcomes. When considered separately, there was not an apparent relationship between the substance of political messages and a variety of political outcomes. In conjunction however, with important features of the source (i.e., manner), relationships were illuminated. For example, perceptions of sender credibility matters. People are capable of reacting or holding values completely opposite from the message they report, when they perceive the source of the message is not credible or dissimilar. Alternatively, people tend to adhere to message recommendations if they perceive that the message sender is similar and/or highly credible. Thus, although the messages collected during this study were not exclusively interpersonal in nature, tenets of Burleson’s message centered approach were supported. Notably understanding the substance and manner of communication allows for a better understanding of outcomes within a given domain. Finally, it is important to consider one of the issues this study encountered when trying to code whether political messages came from an interpersonal or mass media source. Problems of definition have been an issue for interpersonal communication scholars (see Cappella, 1987), and this study provides another example of the difficulty inherent in classifying communication as being one or the other. This is particularly a concern when research collects messages from the population of interest, rather than providing a stimulus message. A growing complication in this 54 area is additionally the ability of technology to mimic face-to-face conversations, especially in the political realm. Topic forums, blogs that are read daily, Twitter, and Facebook, all create opportunities for pseudo-interaction, even though the audience of these messages might be well into the thousands. It is suggested here, that rather than dichotomizing message channels, or relationships, relying on continuous measures of interpersonal may yield further insight into the effect of these messages. The idea that communication channels are not mutually exclusive is not new. Miller and Steinberg (1975) argued over 30 years ago that even face-to-face communication could be viewed as exhibiting various degrees of interpersonal communication. In their seminal book, the authors argue that people’s expectations regarding how to best fulfill their conversational goals dictate whether a conversation is interpersonal in nature. If conversation partners predict how the other will react based on cultural or sociological cues, then Miller and Steinberg would argue that interpersonal communication, in its purest form, did not take place. If however, people interacted and made conversational decisions based on their knowledge of how their partner (at the psychological level) will react, only then has interpersonal communication occurred. Defining mass media communication also suffers from the problem of limited utility, especially with the increased use of technology (now referred to as “new media”). These problems of definition are beyond the scope of the current paper. Regardless, it merits mentioning that these hybrid models of communication are becoming more the rule than the exception, and defining communication channels in such a narrow manner may ultimately hinder a researcher’s ability to understand communication context. Thus, unlike the traditional direction taken in this paper regarding the definitions of mass media and interpersonal, future scholars should consider further 55 refinement of their measures in an effort to understand and identify more completely this new communication environment. Limitations and Future Directions Although several relationships were illuminated in this study, there were also limitations that hindered our ability to make certain claims. The first issue with this study was that college students were the sample of interest. This narrow sample leads to generalizability issues as well as leaves some questions largely unanswered. From a generalizability standpoint, it is unclear whether the current findings would generalize to other voting eligible demographics. College students are unique in terms of: age, an overrepresentation of white participants, and the lack of diversity in educational attainment. These audience features beg the question of whether the same coding category would emerge with a more representative audience. Additionally, due to the limited age range in this sample, it is unclear whether several students recalled quotes from the last election because a) it was recent, and therefore salient, or b) because it was the first election they participated in and therefore salient. An important difference in these two explanations is that if the explanatory mechanism were just a recency effect, than political attitudes and engagement would be a byproduct of the most recent campaign cycle. Alternatively, if memorable messages from one’s first election remain for several years (and elections to come), then political attitudes and behaviors for this generation could be enduring and stable over time. These interesting empirical and mechanistic questions, though yielding important implications, cannot be uncovered with this current dataset. Future work would benefit from the addition of a more represented sample. This representativeness would improve the classification scheme, as well as address whether the recency or primacy effect explains why people recall the memorable messages they do. 56 Another problematic issue in dealing exclusively with the college demographic for political research is that college students tend to be less politically aware and less active (Carpini, 2000) than other portions of the population. This suggests that in this study, more findings might have been uncovered if more variance existed in the political outcome variables assessed. For example, 82 participants failed to answer two or more questions right on the five-item political knowledge test, leading statistically to a restriction in range problem. Furthermore, 35% of the sample participated in two or less of the activities listed on the participation index. Thus, although relationships were uncovered, in the general population there is reason to believe that these effects might be stronger if more variance was built in to the sample. This restriction in range problem was one of the reasons why the engagement index was created, and often yielded significant values. Increasing the variance in this sample did correspond with the uncovering of statistically significant relationships. In addition to increasing the variance, another issue affecting the ability to detect significant findings was statistical power. With ten content categories, a desired sample size to yield effects should be, at a minimum twice as large as the current sample. By increasing sample size, the ability to detect relationships between manner, substance, and outcomes would be improved if relationships actually exist. By virtue of memorable message designs, causal claims are not possible. Causal claims receive support if three design features exist: a) the independent variable precedes the dependent variable in time, b) concomitant variation, and c) alternative explanations are controlled. In this study, only concomitant variation was measurable, timing issues and controls were not possible in this research. Thus, the associations discovered in this manuscript are correlations and not causal in nature. Future research however, can use the different message design types uncovered through this research to pursue how message features can lead to desirable political outcomes. 57 Originally, one concern with the memorable message design was whether Self-Perception theory (Bem, 1965) would explain the relationship between memorable messages and political outcomes if only the predicted relationships were evident. Self-perception theory asserts that people learn information about themselves in the same way they foster impressions about others. They create self-attributions based on observing their own behavior. Thus, if a person voted in the last election, they might report that they are politically active. The implication for this study is that if a person remembered a positive political message, they might answer the political outcome questions as though they are politically engaged. This limitation is inherent in many self-report studies, and consequently here hurts the ability to make causal claims about the impact of substance and manner on tangible outcomes. Although self-perception theory predictions hold in the case of those who recalled a positive message, theoretical propositions fall apart when a negative message is recalled from a politically active participant. For example, self-perception theory would not predict that people recalling a negative political message, would respond positively to political engagement measures. This disconnect provides some support for a causal inference in the case of participants recalling a negative message. This is, however, an empirical question that requires experimental, or longitudinal methods to test. One final limitation discussed here deals with the coding scheme generated by responses. Although most categories were relatively straight-forward, the “non-candidate political quote” category was unfortunately ambiguous. The reason this lack of certainty is troublesome is because participant’s who reported a non-candidate quote tended to rate higher in levels of engagement. Thus, knowing more about this group of people is beneficial. Without more data, 58 and a more detailed account of these types of messages, at this point, not enough is known to make any generalizations or to create message inductions representing this category. Conclusion This research intended to connect message types and message sources to political outcomes. A criticism leveraged against political communication research was the lack of an articulated relationship between the content of effective messages and the impact of message source, on political outcomes. By linking and defining different types of political messages, and demonstrating that their effectiveness depends on perceptions of source credibility and similarity, political communication scholarship is advanced beyond correlational and experimental approaches. Future research is encouraged to consider that all messages are not created equal, and that people’s political behaviors may not just be attributable to a message environment, but rather a specific communicative encounter that inspired political thought and/or action. The relationships uncovered here could have important practical value for educators teaching young students about politics. This research suggests that one way to generate student interest in the subject matter might be to have students debate controversial issues with one another. By allowing students to think through strong and weak arguments originating from referent others, future political activity may be inspired. The relationships between message manner and substance confer status upon the important role of communication specific variables in the political process. These relationships held regardless of whether the messages were interpersonal or mass media in nature. Future scholarship in this vein is encouraged to pursue considering message content and its relationship to political outcomes, especially with a larger and more diverse sample. Additionally, parsing out the effects of message content in conjunction with message source could inform future political message construction aimed at motivating 59 people to become politically active. The findings generated from this work suggest that positive messages from credible sources are just as likely as negative messages from non-credible sources to associate with higher levels of civic engagement. This research also highlighted the importance of considering messages separately rather than focusing exclusively on frequency measures for political talk, or stimulus materials developed for experimental purposes. This methodological extension warrants further investigation in order to examine whether findings using a memorable message paradigm corroborate or contradict the extant of research studying the impact of political communication on political beliefs and behaviors. 60 Notes 1 When inter-rater reliability was calculated using a Scott’s Pi, reliability scores were lower than expected (.44). When looking at the data, this low score is attributable to a misunderstanding between a recall score of 2 or 3. One coder gave substantially more scores of 2 (n = 21) than the other coder (n = 6). In retrospect the difference between a score of 2 or 3 was less objective than the difference between other categories. When participants paraphrased their message, one coder interpreted this paraphrase as indicative of imperfect recall and allotted a 2, whereas the other coder allotted a 3. Given this discrepancy does not influence findings, the mean of coder ratings to represent recall ease was retained despite low categorical reliability ratings. 2 Every notecard contained one message elicited in Phase 1 of the study. For any repeats, meaning that more than one participant recalls a highly similar message (especially likely if a popular political phrase is used), only one notecard containing this message was used. After removing repeats and people who did not report a political message, 51 messages were classified from the original 81. Additionally, all cards were edited so that they contained no grammatical or factual errors. The reason for this was to turn card-sorter’s focus on the content of the message rather than focusing on attributions about the message creator. After cards were sorted, a matrix of dissimilarity was created. Given a total of 51 unique messages were recalled, the size of this matrix was 51 x 51. The entries in the matrix were the number of times the two messages did not get placed in the same cluster (i.e., matrix of noncooccurrence). This matrix was then inputted into SPSS in order to use ALSCAL, which assumes that the input is a dissimilarity matrix. When this matrix was submitted to ALSCAL, a graph helped begin to decipher how the messages clustered. 61 3 In addition to the four-dimensional category solution proffered in the manuscript, a three dimensional solution (see Figure 3) was also ran to examine whether a solution in three dimensions offers more parsimonious conclusions. Results from this analysis were highly similar to the results reported from the four dimensional solution. Specifically, the categories, or clusters, that emerged in both the three and four dimensions were: Obama campaign quotes, campaign slogans from other politicians, party archetypes, September 11th imagery, prodemocracy statements, negative campaigning, non-candidate quotes, issue statements, and unity messages. A category revealed in the three dimensional analysis, that was not found using the four dimensional solution, clustered messages that espoused a personal opinion. Unlike in the four dimensional solution, however, 10 messages remained that could not be easily placed in a category because it did not fit neatly into a cluster (e.g., M16 in Figure 3). When running this data in four-dimensions however, the distance between the objects did allow all messages to be placed more confidently into categories. For this reason, and due to the interpretive similarities in both solutions, the original four-dimensional solution was retained. 62 Appendices 63 Appendix A Memorable Message Prompt This study is interested in memorable messages that you recall about politics. Individuals receive hundreds of messages each day, yet most are simply processed and forgotten. However, there may be a few messages, which are called MEMORABLE MESSAGES, that you recall for long periods of time and which you may think about in different contexts. Remembering these precise messages affects your behavior, your attitudes, and may even help you make sense of the world. In some cases, you may have heard the message many times but it did not become memorable until it was said in a particular context by a particular person. In other cases, the message may have had a lasting impact the first time you heard it. This survey is particularly interested in memorable messages that best represent what you think about, when you think about politics. Politics has a different meaning for different people. Thus, when people think about politics they can think about a variety of different political messages they have heard. In this study, we want to know what message you think about when you think about politics. This message can come from anywhere. For example, some people may recall political messages they received from a person they know such as a parent, relative or friend. On the other hand, some people will think about a message they received from the mass media, such as a message from a political interview, speech, TV show, etc. For this study, we would like you to recall a message that comes to mind when you think about politics. You could have heard this message multiple times, but the key is that it served as a guide for how you think about politics, as well, perhaps, a guide to your political behavior. We ask that you precisely recall the words of your memorable political message. Take a few minutes to think about a memorable message that you have received that guides your understanding of politics. Once you have recalled that message, please answer the questions that follow. 64 Appendix B Survey with Scale Labels Memorable Message Content • What is the message? Please record the message word-for-word. Message Source • • • • • • • Who was the source of this message the first time you heard it? Do you know the source of your message on a personal level? Have you communicated with this person about other topics in the past? Was this message directed specifically towards you? Did this message occur as a part of a longer discussion? Would this person know you if you ran into them on the street? How did you receive this message? face to face discussion, phone, email, instant message device, through texting, TV, radio, website, or other. Message Valance • • • What emotions did or do you feel when you recall this message? Overall, would you consider this message to be: positive negative neutral Positive (1) 2 3 4 5 6 (7) Negative Politically Pessimistic (1) 2 3 4 5 6 (7) Politically Optimistic Favorable (1) 2 3 4 5 6 (7) Unfavorable Encouraging (1) 2 3 4 5 6 (7) Discouraging Good (1) 2 3 4 5 6 (7) Bad Source’s Political Credibility & Similarity Please use the scales below to classify your perceptions regarding the source of your message when it comes to their political opinions: Experienced (1) 2 3 4 5 6 (7) Inexperienced Uninformed (1) 2 3 4 5 6 (7) Informed Expert (1) 2 3 4 5 6 (7) Non-expert 65 Biased (1) 2 3 4 5 6 (7) Non-biased Honest (1) 2 3 4 5 6 (7) Dishonest Trustworthy (1) 2 3 4 5 6 (7) Untrustworthy Open-Minded (1) 2 3 4 5 6 (7) Close-Minded Unjust (1) 2 3 4 5 6 (7) Just Fair (1) 2 3 4 5 6 (7) Unfair Unselfish (1) 2 3 4 5 6 (7) Selfish Please answer these next questions about your level of similarity to the message source in the following domains. Political Affiliation Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Interpersonal Style Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Political Interests Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Religious Beliefs Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Income Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Age Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Political Knowledge Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar General Attitudes Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Education Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Personality Very Dissimilar (1) 2 3 4 5 6 (7) Very Similar Political Engagement – Dependent Variables Political Participation • • • • Did you vote in the last National election? Have you ever voted in a Local election? Did you do any formal work for the 2008 Presidential campaign? Did you do any formal work for the ‘07-’08 Primary campaign? 66 • • • • • • • • Have you ever attended a political rally? Have you ever attended a political protest? Have you ever signed a petition? Have you ever been involved in a university sponsored political organization? Have you ever been involved in any university sponsored organization? Have you ever contacted a politician? Have you ever visited a political campaign website? Have you ever donated money to a political campaign? Political Interest • • • • • How would you characterize your political affiliation? Where would you place yourself on this scale? Very Liberal (1) to Very Conservative (7) How would you rate your political knowledge? Poor (1) – Excellent (7) How confident are you in your ability to recall political facts and information? Not at all (1) – Very (7) How interesting is politics to you? Not at all (1) – Very (7) Political Knowledge • • • • • What European country recently received a major bailout from the European Union? Who is the Chief Justice of the Supreme Court? Who is currently the Vice President? How long is a U.S. Senator’s term in office? What U.S. President was in office during the Bay of Pigs invasion? 67 Appendix C Instructions for Phase 2 Thank you for your participation. The purpose of this study is to help us sort messages into appropriate categories. In order to do so, we are having students come in and sort cards into categories based on content similarity. When numerous students independently agree on a category, as researchers we can infer that this category is valid. Therefore today, we are asking you to take the set of notecards on the table, and pile them into categories based on similarity. You may create as many categories as you want, however, each card can only be in one category. After you have done this, please fill out some brief information about each of the categories, so we have an understanding of how you arrived at the piles you did. Do you have any questions? Description of Each Pile What characteristic do cards in this pile share? 68 Appendix D Researcher Generated Source Role Coding Scheme Category Name Politician, Ex-Politician or Politician’s Spouse Category Descriptors • • • • Pundit • • • Family Member • Friends • Artist • Activist • • • • • • Category Examples Message that come directly from a politician This can include expoliticians as well Local and National level politicians at all levels of government Also count politicians who ran for office but might not have won A journalist Anyone who works for a news network (e.g, CNN, FOX) A person who primarily discusses political matters Anyone who is related to the person who stated the message Messages coming from a person who is not related to the person • • • • Barack Obama John Kerry Sarah Palin John F. Kennedy • • • • • Rush Limbaugh John Stewart Anderson Cooper Bill O’Reilly Thomas Friedman • • • • Mom, dad, grandparent, sibling, cousin, aunt, uncle, etc Friend Family friend Classmate, coworker, boss Messages coming from a person who is in the public eye, but not necessarily for their political opinions Actors or actresses Musicians Comedians A person who is vocal about their political beliefs but does not and has never held a political office. This person is also NOT employed by any news station or newspaper This person is well-known solely for their political activism • • • • Sean Penn Jay-Z Daniel Tosh Oprah • Martin Luther King Jr. 69 Unknown Person • • Political Advertisement • Propaganda • Political Material • • • Other • • A person in mentioned but they cannot be identified Not a well known politician, pundit, activist, or artist An advertisement that is related to a political campaign but is not directly associated with the campaign A bumper sticker or lawn sign A t-shirt, graffiti, or poster A political pin A textbook, film or documentary, class material • “A business man sitting at the table next to me at a restaurant” • “A political advertisement” • “A poster in my high school” • “A documentary in my ISS class my freshman year” Anything that does not fall • into the above categories Includes when a referent was so vague that it could not be sorted into any of the above categories 70 “TV” Table 1. Researcher generated coding scheme categories compared with card sorting coding categories, along with number of messages falling into each category and number of interpersonal versus mass communicated messages. Message Content Category Message Example Media n Interpersonal n Research Generated Coding Scheme Campaign Slogans “Change we can believe in.” 54 2 Negative Messaging “Politics are mostly a bunch of malarkey and don’t really care about the people” 19 3 Pro-Democracy Sentiments “It is your right and obligation as an American citizen to vote.” 26 3 Political Archetypes “Democrats are looking for handouts or are usually young and trying to save the world” 4 8 3 2 America’s Divide “A house divided against itself cannot stand” Non-Candidate Quotes “People should not be afraid of their governments, governments should be afraid of their people.” 13 3 Political Events “Ladies and gentleman, I would like to introduce to you the President of the United States… Barack Obama” 25 3 12 3 Political Issues “A police chief’s stance on the war on drugs. How we are losing and should devote time and money elsewhere.” 71 Table 1 (cont’d) Student Generated Coding Scheme Non-Obama Slogans “Make Michigan work again.” 11 0 Negative Campaigning “I don’t remember one in particular, I just remember how annoyed I am with all the political ads speaking poorly of their opponents.” 21 4 Pro-Democracy Sentiments “Ask not what your country can do for you, but what you can do for your country.” 14 2 “In general my values align with those typical of a Democrat, so I will always vote democratically.” 5 10 9 3 14 2 Political Archetypes Non-Candidate Quotes Political Events Political Issues “Don’t tread on me.” “When I attended ASU, Obama came and spoke before his election. I don’t remember the specific speech.” “George W. Bush talking about helping kids.” 13 0 Uniting Messages “You always need to understand both sides of an issue before you form an opinion about it.” 17 1 Non-Domestic Affairs “How the people in Egypt have the right for speaking up and” asking for their rights in their country.” 5 2 9/11 Events “The way George Bush reacted to the 911 tragedy, stating that he would not their actions go unpunished.” 7 1 43 2 Obama Slogans “Yes we can” 72 Table 2. Civic engagement means by all memorable message categories. Message Content Political Knowledge Political Participation M (SD) M (SD) M (SD) 2.14 (1.38) 1.25 (1.28) 1.89 (1.33) 1.80 (1.62) 2.25 (0.96) 2.29 (1.80) 2.31 (1.40) 2.00 (3.53) 4.72 (3.14) 4.38 (4.17) 5.06 (2.39) 3.10 (3.04) 4.00 (1.41) 4.71 (1.60) 3.75 (2.38) 5.50 (2.88) 3.32 (1.44) 3.08 (1.15) 3.51 (1.58) 3.17 (1.61) 2.67 (1.19) 3.67 (1.20) 3.13 (1.32) 3.28 (1.14) 2.00 (1.61) 1.96 (1.37) 2.00 (1.21) 1.73 (1.28) 2.42 (1.38) 1.75 (1.57) 1.54 (1.56) 1.72 (1.27) 1.43 (1.13) 2.88 (1.46) 2.22 (1.41) 4.73 (3.04) 3.64 (3.17) 3.44 (2.39) 4.00 (3.07) 5.08 (2.19) 3.38 (2.66) 3.92 (3.35) 5.00 (2.62) 3.29 (1.98) 4.38 (2.39) 4.04 (2.78) 3.18 (0.98) 3.21 (1.39) 3.21 (0.88) 2.82 (1.19) 3.42 (1.39) 3.21 (1.41) 2.92 (1.40) 3.22 (1.60) 2.90 (1.10) 3.00 (1.18) 3.32 (1.38) Category Political Interest Engagement Scale M (SD) Research Generated Coding Scheme Campaign Slogans Negative Messaging Pro-Democracy Sentiments Political Archetypes America’s Divide Non-candidate Quotes Political Events Political Issues 10.18 (4.00) 8.71 (5.27) 10.65 (3.86) 8.07 (4.91) 8.92 (2.73) 10.67 (3.73) 9.19 (3.70) 10.78 (3.49) Student Generated Coding Scheme Non-Obama Slogans Negative Campaigning Pro-Democracy Sentiments Political Archetypes Non-candidate Quotes Political Events Political Issues Uniting Messages Non-Domestic Affairs 9/11 Events Obama Slogans 73 9.91 (4.59) 8.81 (4.06) 8.65 (3.60) 8.56 (4.09) 10.92 (3.99) 8.33 (3.89) 8.38 (4.10) 10.12 (3.90) 7.62 (2.59) 10.25 (3.74) 9.59 (3.77) Table 3. Predicting civic engagement by message valance and source credibility. Knowledge Predictors B (SE) Participation b B (SE) Interest b b b 3.48 (1.10) Valance -0.08 (.22) -.10 -.69 (.45) -.40 -.65 (.20) -.82** -1.48 (.63) -.61* Credibility -0.37 (.25) -.31 -.49 (.51) -.20 -.36 (.23) -.32 -1.26 (.71) -.36 0.03 (.05) .27 .14 (.09) .58 .11(.04) R 5.45 (1.01) B (SE) Intercept Valance X Credibility 6.46 (2.22) B (SE) Engagement Scale 15.56 (3.11) 1.01** .16 Adj. R 2 .13 .25 .17 .01 .00 .05 .01 Note: The interaction term should be interpreted as the cross-product between these two variables. T .29 (.13) p < .05, one tailed, *p < .05, ** p < .01 74 T .88* Table 4. Correlations between message valance and civic engagement variables under conditions of low, medium, and high source expertise. Low Credibility Credibility r r (n = 50) Civic Engagement Medium High (n = 72) Credibility r (n = 55) Political Knowledge -.13 .02 .18b Political Participation -.14 .02 .10b Political Interest -.46a** -.11 .02b General Engagement -.29a* -.01 .19b Note: Due to the way valance was measured, a negative correlation can interpreted such that as messages become more positive, civic engagement increases. Subscripts indicate that correlations are significantly different from one another at p < .05. * p < .05, ** p < .01 75 Table 5. Multiple regression equations examining whether message valance, source credibility, and source similarity predict levels of political engagement. Knowledge Intercept 3.36 (2.80) Valance .16 (.63) .18 -1.72 (1.26) -1.01 -.69 (.56) -.87 -2.32 (1.72) -.97 Credibility -.15 (.63) -.12 -1.75 (1.26) -.71 -.77 (.56) -.67 -2.71 (1.72) -.78 Similarity .48 (.38) .43 .46 (.76) .20 .73 (.34) .69* 1.61 (1.03) .51 VxC -.06 (.13) -.40 .27 (.25) 1.15 .08 (.11) .71 .31 (.34) .94 VxS -.11 (.12) -.51 .08 (.23) .20 -.04 (.10) -.18 -.05 (.32) -.09 CxS -.11 (.13) -.38 .02 (.27) .04 -.13 (.12) -.49 -.20 (.37) -.26 .02 (.03) .38 -.06 (.07) -.63 -.00 (.03) -.10 -.05 (.09) -.35 2 2 .04 (.00) b B (SE) Total Engagement B (SE) R (Adj. R ) B (SE) Interest Predictors VxCxS b Participation 13.66 (5.60) b 7.75 (2.49) .05 (.01) .14 (.10) 76 B (SE) b 24.91 (7.67) .09 (.05) Table 6. Correlations between message valance and political engagement variables under conditions of low, medium, and high source credibility and similarity. Low Credibility Civic Engagement Low Similarity Medium Credibility r r (n = 24) High Credibility r (n = 24) (n = 17) Political Knowledge -.08 -.27 .27b Political Participation -.23 -.12 .05 Political Interest -.52a** -.31a .15 General Engagement -.36 a -.27 .18 Moderate Similarity (n = 18) (n = 28) (n = 16) Political Knowledge -.17 .00 -.12 Political Participation .04 .09 .39b Political Interest -.34a -.05 -.03 General Engagement -.14 .05 .24 High Similarity (n = 8) Political Knowledge -.27 .38b .42b* Political Participation -.19 .18 -.12 Political Interest -.73a* .02 .11 General Engagement -.44a .26 .09 T (n = 20) 77 (n = 22) Note: Different subscripts denote that correlations are significantly different from one another at p < .05 using a Fishers R-to-Z test of correlation differences. 78 Figure 1. Mean plot portraying the influence of message valance and source credibility on participants’ political interest. Note: None of these mean differences are significant at the p < .05 level. : 79 Figure 2. Mean plot portraying the influence of message valance and source credibility on source engagement index scores. Note: None of these mean differences are significant at the p < .05 level. : 80 Figure 3. A visual depiction of the three dimensional solution obtained using multidimensional scaling with ALSCAL. Note: The labels next to each point represent the message associated with that location on the els multidimensional scale. 81 References 82 REFERENCES Ansolabehere, S., & Iyengar, S. (1995). Going negative: How attack ads shrink and polarize the electorate. New York: Free Press. Bailenson, J. N., Garland, P., Iyengar, S., & Lee, N. (2006). Transformed facial similarity as a political cue: A preliminary investigation. Political Psychology, 27, 373-385. Bailey, C. H., & Chen, M. (1989). Time course of structural changes at identified sensory neuron synapses during long-term sensitization in aplysia. Journal of Neuroscience, 9, 1774-1781. Beck, P. A., & Jennings, M. K. (1991). Family traditions, political periods, and the development of partisan orientations. Journal of Politics, 53, 742-763. Bem, D. J. (1965). An experimental analysis of self-persuasion. Journal of Experimental Social Psychology, 1, 190-218. Burleson, B. (2009). The nature of interpersonal communication: A message-centered approach. In C. R. Berger, M. E. Roloff, & D. R. Roskos-Ewoldson (Eds.), The handbook of communication science 2nd ed. (pp. 145-163). Thousand Oaks, CA: Sage. Capella, J. N. (1987). Interpersonal communication: Definitions and fundamental questions. In C. R. Berger & S. H. Chaffee (Eds), Handbook of Communication Science (pp. 184-238). Newbury Park, CA: Sage. Carpini, M. X. D. (2000). Gen.com: Youth, civic engagement, and the new information environment. Political Communication, 17, 341-349. Chaiken, S. (1980). Heuristic versus systematic information processing and the use of source versus message cues in persuasion. Journal of Personality and Social Psychology, 39, 752-766. Chambers, S. (2003). Deliberative democratic theory. Annual Review of Political Science, 6, 307-326. Cohen, G. L. (2003). Party of policy: The dominating impact of group influence on political beliefs. Journal of Personality and Social Psychology, 85, 808-822. Conover, P. J., & Searing, D. D. (2005). Studying ‘everyday political talk’ in the deliberative system. Acta Politica, 40, 269-283. Conover, P. J., Searing, D. D., & Crewe, I. M. (2002) The deliberative potential of 83 political discussion. Journal of Political Science, 32, 21-62. Converse, P. (1964). The nature of belief systems in mass publics. In D. Apter (Ed.), Ideology and discontent. New York, NY: Free Press. Deaux, K., Reid, A., Mizrahi, K., & Ethier, K. A. (1995). Parameters of social identity. Journal of Personality and Social Psychology, 68, 280-291. Carpini, M. X., Cook, F. L., & Jacobs, L. R. (2007). Public deliberations, discursive participation and citizen engagement: A review of empirical literature. Annual Review of Political Science, 7, 315-344. Carpini, M. X., & Keeter, S. (1997). What Americans know about politics and why it matters. New Haven, CT: Yale University Press. Carpini, M. X., Keeter, S. (1993). Measuring political knowledge: Putting first things first. American Journal of Political Science, 37, 1179-1206. Downs, A. (1957). An economic theory of political action in a democracy. The Journal of Political Economy, 65, 135-150. Druckman, J. N. & Nelson, T E. (2003). Framing and deliberation: How citizens’ conversations limit elite influence. American Journal of Political Science, 47, 728-744.Ellis, J. B., & Smith, S. W. (2004). Memorable messages as guides to self-assessment of behavior: A replication and extension diary study. Communication Monographs, 71, 97-119. Fishkin, J., & Luskin, R. (1999). A poll with a human face: National Issues Convention experiment in political communication. In M. McCombs and A. Reynolds (Eds.), Bringing deliberation to the democratic dialogue: The NIC and beyond, (pp. 30-38). New York: Erlbaum. Ford, L. A., & Ellis, B. H. (1998). A preliminary analysis of memorable support and nonsupport messages received by nurses in acute care settings. Health Communication, 10, 37-63. Fowler, J. H. (2005). Turnout in a small world. In A. Zuckerman (Ed), Social logic of politics, (pp. 269-287). Philadelphia: Temple University Press. Halladay, S. (2002). “Have fun while you can,” “You are only as old as you feel,” and “Don’t ever get old!”: An examination of memorable messages about aging. Journal of Communication, 52, 681-697. Hamilton, M. A., & Hunter, J. E. (1988). CFA 1.8 [Computer Software]. Storrs, CT: University of Connecticut. 84 Hanisch, C. (1969). The personal is political. In S. Firestone and A. Koedt (Eds.), Notes from the second year: Women’s liberation. Publisher unknown. Hovland, C. I., Janis, I. L., & Kelley, J. J. (1953). Communication and persuasion. New Haven, CT: Yale University Press. Huckfeldt, R., & Sprague, J. (1991). Discussant effects on vote choice: Intimacy, structure, and interdependence. Journal of Politics, 53, 122-158. Jennings, M. K., & Niemi, R. G. (1981). Generations and politics. Princeton, NJ: Princeton University Press. Kassing, J. W., & Pappas, M. E. (2007). Champions are built in the off season, Human Communication, 10, 537-546. Katz, E., & Lazarsfeld, P. F. (1955). Personal influence: The part played by people in the flow of mass communications. Glencoe, IL: The Free Press. Kellermann, K., Reynolds, R., & Chen, J. B. (1991). Strategies of conversational retreat: When parting is not sweet sorrow. Communication Monographs, 58, 362 – 383. Kim, J., Wyatt, R., & Katz, E. (1999). News, talk, opinion, and action: The part played by conversation in deliberative democracy. Political Communication, 16, 361-385. Klofstad, C. A. (2009). Civic talk and civic participation: The moderating effect of individual predispositions. American Politics Research, 37, 856-878. Knapp, M. L., Stohl, C., & Reardon, K. K. (1981). Memorable messages. Journal of Communication, 31, 27-41. Lau, R. R., Sigelman, L., Heldman, C., & Babbit, P. (1999). The effects of negative political advertisements: A meta-analytic assessment. American Political Science Review, 93, 851-875. Mackie, G. (August, 2002). Does deliberation change minds? Paper presented at the American Political Science Association Annual Conference in Boston, MA. Medved, C. E., Brogan, S. M., McClanahan, A. M., Morris, J. F., & Shepherd, G. J. (2006). Family and work socializing communication: Messages, gender, and ideological implications. The Journal of Family Communication, 6, 161-180. Miller, G. R., & Steinberg, M. (1975). Between people: A new analysis of interpersonal communication. Chicago: Science Research Associates. Mutz, D. C. (2002). The consequences of cross-cutting networks for political participation. American Journal of Political Science, 46, 838-855. 85 Mutz, D. C. (2006). Hearing the other side: Deliberative versus participatory democracy. New York: Cambridge University Press. Nickerson, D. W. (2008). Is voting contagious? Evidence from two field experiments. American Political Science Review, 102, 49-57. Noelle-Neumann, E. (1974). The spiral of silence: A theory of public opinion. Journal of Communication, 24, 43-51. O’Keefe, D. J. (2002). Persuasion theory & research. Thousand Oaks, CA: Sage Publications. O’Keefe, D. J. (2003). Message properties, mediating states, and manipulation checks: Claims, evidence, and data analysis in experimental persuasive message effects research. Communication Theory, 13, 251-274. Petty, R.E., Cacioppo, J.T., & Goldman, R. (1981). Personal involvement as a determinant of argument-based persuasion. Journal of Personality and Social Psychology, 41, 847 855. Price, V., Nir, L., Capella, J., (2002). Does disagreement contribute to more deliberative opinion?. Political Communication, 19, 95-112. Robberson, M. R., & Rogers, R. W. (1988). Beyond fear appeals: Negative and positive persuasive appeals to health and self-esteem. Journal of Applied Psychology, 18, 277-287. Robinson, J. P. (1976). Interpersonal influence in election campaigns: Two step flow hypotheses. Public Opinion Quarterly, 40, 301-319. Rosenberg, S. W. (1988). Reason, ideology, and politics. Princeton, NJ: Princeton University Press. Schacter, D. L. (1999). The seven sins of memory: Insights from psychology and cognitive neuroscience. American Psychologists, 54, 182-203. Shulman, H. C., Boster, F. J., & Carpenter, C. (November, 2010). The effect of test-taking venue and test format on political knowledge test performance. Paper accepted for presentation at the annual meeting of the National Communication Association, San Francisco, CA. Shulman, H. C., Smith, S. W., & Clark-Hitt, R. (unpublished manuscript). An investigation into political socialization processes within the family. 86 Shulman, H. C., & Wittenbaum, G. M. (November, 2010). Promoting perceived deliberative success through group communication. Paper to be presented at the National Communication Association annual meeting in San Francisco, CA. Sigelman, L., & Sigelman, C. K. (1982). Sexism, racism, and ageism in voting behavior: An experimental analysis. Social Psychology Quarterly, 45, 263-269. Smith, S. W., & Ellis, J. B. (2001). Memorable messages as guides to self-assessment of behavior: An initial investigation. Communication Monographs, 68, 154-168. Smith, S. W., Ellis, J. B., & Yoon, H. (2001). Memorable messages as guides to selfassessment of behavior: The role of instrumental values. Communication Monographs, 68, 154-168. Smith, S. W., Munday, S., LaPlante, C., Kotowski, M. R., Atkin, C., Skubisz, C. M., & Stohl, C. (2009). Topics and sources of memorable breast cancer messages and their impact on prevention and detection behaviors. Journal of Health Communication, 14, 293-307. Stimson, J. A. (2004). Tides of consent: How public opinion shapes American politics. New York: Cambridge University Press. Stohl, C. (1986). The role of memorable messages in the process of organizational socialization. Communication Quarterly, 34, 231-249. Stoker, L. (August, 2007). Growing up with politics. Paper presented at the annual meeting of the American Political Science Association in Chicago, IL. Sunstein, C. (2002). The law of group polarization. Journal of Political Philosophy, 10, 175- 95. Tetlock, P. E. (1986). A value pluralism model of ideological reasoning. Journal of Personality and Social Psychology, 50, 819-827. Verba, S, Schlozman, K. L., & Brady, H. (1995). Voice and Equality: Civic Voluntarism in American Politics. Cambridge, MA: Harvard University Press. Westen, D. (2007). The political brain: The role of emotion in deciding the fate of a nation. New York: PublicAffairs. Wyatt, R. O., Katz, E., & Kim, J. (2000). Bridging the spheres: Political and personal conversation in public and private spaces. Journal of Communication, 50, 71-92. Zaller J., & Feldman, S. (1992). A simple theory of survey response: Answering Questions versus revealing preferences. American Journal of Political Science, 87 36, 579-616. Zaller. J. R. (1992). The nature and origins of mass opinion. New York, NY: Cambridge University Press. 88