The impacts of visual message features in cancer risk communication

Visuals are often used to communicate health information, including information about cancer risk. This dissertation launched a research program by conducting three studies utilizing multiple methodological perspectives. These three studies collectively aimed to understand the impact of using visuals in communicating cancer risk. The first study-a meta-analysis of visual message experiments-summarized six studies (eight effect sizes) involving 5,382 participants and demonstrated a significant positive effect of using visuals to alter cancer risk perception (Cohen's d = .10, 95% CI: .05, .16). The second study is a 5 (visual: UV skin damage, sun exposure, sunburn, photoaging, and mole removal) × 3 (replication: three examples of each visual condition) × 4 (efficacy: no efficacy, text only, visual, visual + text) randomized controlled trial. The study illustrates that UV skin damage visuals, when compared to all other visual conditions combined, generated greater fear, which then triggered greater sun-safe behavior expectations. When compared with other visual conditions separately, only mole removal visuals produced equivalent fear as UV skin damage visuals. The third study followed up the second study in two important ways: (1) adding a personalized UV (PUV) photo condition and (2) utilizing physiological measures (i.e., facial expression and skin conductance). Results showed that PUV skin damage photos produced significantly greater self-reported fear and positive valence (detected by facial expression analysis) than NUV but not SUV. Mediation analysis demonstrated that novelty iv significantly mediated the relationship between exposure to PUV vs. NUV (and SUV vs. NUV) skin damage visuals and all three types of behavior expectations (i.e., sunscreen, protective items, and seeking shade). Overall, this dissertation made contributions to visual health communication in the context of cancer. Chapters 3 and 4 added to the scholarly conversation by investigating the mediating role of fear and other discrete emotions. The dissertation found critical differences between personalized UV photos, stock UV photos, and mole removal photos. Other important findings from the program relate to the importance of novelty in messages. When messages are perceived as novel, they are more effective in influencing behavioral outcomes.

THE IMPACTS OF VISUAL MESSAGE FEATURES IN CANCER RISK COMMUNICATION by Manusheela Pokharel A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Communication The University of Utah August 2019 Copyright © Manusheela Pokharel 2019 All Rights Reserved The University of Utah Graduate School STATEMENT OF DISSERTATION APPROVAL Manusheela Pokharel The dissertation of has been approved by the following supervisory committee members: Jakob D. Jensen , Chair Kimberly A. Kaphingst , Member Ye Sun , Member Sara K. Yeo , Member Lee Ellington , Member and by Danielle Endres the Department/College/School of and by David B. Kieda, Dean of The Graduate School. 4/10/2019 Date Approved 4/10/2019 Date Approved 4/10/2019 Date Approved 4/10/2019 Date Approved 4/10/2019 Date Approved , Chair/Dean of Communication ABSTRACT Visuals are often used to communicate health information, including information about cancer risk. This dissertation launched a research program by conducting three studies utilizing multiple methodological perspectives. These three studies collectively aimed to understand the impact of using visuals in communicating cancer risk. The first study—a meta-analysis of visual message experiments—summarized six studies (eight effect sizes) involving 5,382 participants and demonstrated a significant positive effect of using visuals to alter cancer risk perception (Cohen’s d = .10, 95% CI: .05, .16). The second study is a 5 (visual: UV skin damage, sun exposure, sunburn, photoaging, and mole removal) × 3 (replication: three examples of each visual condition) × 4 (efficacy: no efficacy, text only, visual, visual + text) randomized controlled trial. The study illustrates that UV skin damage visuals, when compared to all other visual conditions combined, generated greater fear, which then triggered greater sun-safe behavior expectations. When compared with other visual conditions separately, only mole removal visuals produced equivalent fear as UV skin damage visuals. The third study followed up the second study in two important ways: (1) adding a personalized UV (PUV) photo condition and (2) utilizing physiological measures (i.e., facial expression and skin conductance). Results showed that PUV skin damage photos produced significantly greater self-reported fear and positive valence (detected by facial expression analysis) than NUV but not SUV. Mediation analysis demonstrated that novelty significantly mediated the relationship between exposure to PUV vs. NUV (and SUV vs. NUV) skin damage visuals and all three types of behavior expectations (i.e., sunscreen, protective items, and seeking shade). Overall, this dissertation made contributions to visual health communication in the context of cancer. Chapters 3 and 4 added to the scholarly conversation by investigating the mediating role of fear and other discrete emotions. The dissertation found critical differences between personalized UV photos, stock UV photos, and mole removal photos. Other important findings from the program relate to the importance of novelty in messages. When messages are perceived as novel, they are more effective in influencing behavioral outcomes. iv Dedicated to Buba (My Father-in-Law) The Man of Wisdom. It’s a visual world and people respond to visuals. Joe Sacco TABLE OF CONTENTS ABSTRACT....................................................................................................................... iii LIST OF TABLES .............................................................................................................. x LIST OF FIGURES ........................................................................................................... xi ACKNOWLEDGMENTS ................................................................................................ xii Chapters 1. INTRODUCTION .......................................................................................................... 1 1.1 Dissertation Chapter Outline................................................................................ 3 1.2 Cancer .................................................................................................................. 6 1.2.1 Cancer Risk Communication ................................................................... 6 1.3 Visual Research ................................................................................................... 7 1.4 Skin Cancer Prevention........................................................................................ 8 1.4.1 UV Photo Intervention ............................................................................. 9 1.5 Theory Overview and Selection......................................................................... 10 1.5.1 Potential Theoretical Frameworks ......................................................... 11 1.5.2 Theory to Guide UV Research ............................................................... 14 1.6 Implications in Communication Literature ........................................................ 15 1.6.1 Advance Understanding of EPPM ......................................................... 15 1.6.2 Incorporates Physiological Measures in Existing Theoretical Framework ...................................................................................................... 16 1.6.3 Advance Understanding of Visual Message Features ........................... 16 1.6.4 Investigates the Utility of UV Photographs in Promoting Sun-safe Behaviors ........................................................................................................ 17 1.6.5 Future Implications in Information Processing Literature ..................... 17 1.7 Conclusion ......................................................................................................... 18 2. THE IMPACT OF VISUALS ON CANCER RISK PERCEPTIONS: A METAANALYSIS ....................................................................................................................... 20 2.1 Visual Health Communication ........................................................................... 20 2.2 Moderators ......................................................................................................... 22 2.2.1 Visual Format......................................................................................... 22 2.2.2 Sample Characteristics ........................................................................... 23 2.2.3 Sampling Method ................................................................................... 23 2.3 Method ............................................................................................................... 24 2.3.1 Article Identification and Selection ....................................................... 24 2.3.2 Unit of Analysis ..................................................................................... 25 2.3.3 Coding .................................................................................................... 25 2.3.4 Extracting Effect Sizes ........................................................................... 26 2.4 Results ................................................................................................................ 27 2.4.1 Overall Effect Size ................................................................................. 27 2.4.2 Test for Heterogeneity ........................................................................... 28 2.5 Discussion .......................................................................................................... 28 2.6 Conclusion ......................................................................................................... 30 3. DO ULTRAVIOLET PHOTOS INCREASE SUNSAFE BEHAVIOR EXPECTATIONS VIA FEAR? A RANDOMIZED CONTROLLED TRIAL IN A SAMPLE OF U.S. ADULTS ........................................................................................... 36 3.1 Introduction ........................................................................................................ 36 3.2 The EPPM: A Framework for Studying Fear Appeals ...................................... 38 3.2.1 Skin Cancer Prevention, Ultraviolet (UV) Photo Interventions, and the EPPM .............................................................................................................. 39 3.3 Method ............................................................................................................... 44 3.3.1 Participants and Procedure ..................................................................... 44 3.3.2 Study Design .......................................................................................... 44 3.3.3 Stimuli .................................................................................................... 44 3.3.4 Measures ................................................................................................ 46 3.3.5 Randomization Check ............................................................................ 49 3.3.6 Power Analysis ...................................................................................... 50 3.4 Results ................................................................................................................ 50 3.4.1 Bivariate Correlations ............................................................................ 50 3.4.2 UV vs. All (H1 & H2) ........................................................................... 51 3.4.3 Fear as a Mediator (H3) ......................................................................... 52 3.4.4 UV Efficacy as a Moderator (H4) .......................................................... 52 3.4.5 Comparison of UV to Other Naturally Occurring Visual Categories (RQ1) .............................................................................................................. 53 3.5 Discussion .......................................................................................................... 54 3.6 Supplemental Materials ..................................................................................... 67 4. VISUAL TAILORING AND SKIN CANCER PREVENTION: COMPARING PERSONALIZED, STOCK, AND NON-ULTRAVIOLET IMAGES ........................... 82 4.1 Studying Fear Appeals ....................................................................................... 84 4.1.1 Skin Cancer Prevention, Ultraviolet (UV) Photo Interventions, and the EPPM .............................................................................................................. 85 4.2 Visual Tailoring ................................................................................................. 86 4.3 Emotion and Physiological Methods in Fear Appeal Research ........................ 89 4.4 Methods.............................................................................................................. 91 4.4.1 Study Design .......................................................................................... 91 4.4.2 Participants and Procedure ..................................................................... 91 4.4.3 Stimuli .................................................................................................... 93 viii 4.4.4 Measures ................................................................................................ 93 4.4.5 Randomization Check ............................................................................ 98 4.4.6 Power Analysis ...................................................................................... 99 4.5 Results ................................................................................................................ 99 4.5.1 Bivariate Correlations ............................................................................ 99 4.5.2 Comparing PUV, SUV, and NUV (H1, H2, H4) ................................. 100 4.5.3 Fear and Physiological Arousal as Mediators (H3, H5) ...................... 101 4.5.4 Positive and Negative Valence (RQ1 and RQ2) .................................. 102 4.5.5 The Role of Novelty............................................................................. 103 4.6 Discussion ........................................................................................................ 103 4.7 Supplemental Materials ................................................................................... 121 5. CONCLUDING REMARKS ...................................................................................... 125 5.1 Contributions.................................................................................................... 128 5.1.1 Visual Features and Impact .................................................................. 128 5.1.2 Affective Constructs as a Mediator...................................................... 129 5.1.3 Evidence Base for Cancer Communication and Control .................... 129 5.2 Moving Forward .............................................................................................. 130 5.2.1 Future Studies ...................................................................................... 131 5.3 Conclusion ....................................................................................................... 136 REFERENCES ............................................................................................................... 138 ix LIST OF TABLES Tables 2.1 Codebook ................................................................................................................... 31 2.2 Variables, Operational Definitions, and Intercoder Reliability .................................. 32 2.3 Study Effect Sizes, Variance, Standard Error ............................................................. 33 3.1 Bivariate Correlations ................................................................................................. 61 3.2 Means and Standard Errors for Mediators and Outcomes by Visual and Efficacy Condition........................................................................................................................... 62 3.3 Simple Mediation – Tests of Indirect Effects of EPPM Variables ............................. 63 3.4 Means and Standard Errors of Mediators and Outcomes by Visual Condition .......... 64 4.1 Bivariate Correlations ............................................................................................... 109 4.2 Means and Standard Errors for Mediators and Outcomes by Visual and Efficacy Condition......................................................................................................................... 110 4.3 Means and Standard Errors for Physiological Indicators (in time percent) by Visual and Efficacy Condition ................................................................................................... 111 LIST OF FIGURES Figures 2.1 Screening flowchart .................................................................................................... 34 2.2 Effect size distribution ............................................................................................... 35 3.1 Simple mediation model of the EPPM........................................................................ 65 3.2. Moderated mediation – indirect effect at four levels of efficacy condition. .............. 66 4.1 Simple mediation models (Fear as mediator) ........................................................... 112 4.2 Simple mediation models (Physiological Arousal as mediator) ............................... 114 4.3 Simple mediation models (Positive Valence as mediator) ....................................... 116 4.4 Simple mediation models (Novelty as a mediator) ................................................... 118 ACKNOWLEDGMENTS My quest for getting a Ph.D. started when I was very young. My dad’s academic life has been instrumental in shaping my life trajectory. I remember looking at his graduation picture and being deeply influenced. I would stare at that picture and try to imagine myself in it and was determined to do whatever it took to achieve that. The completion of this dissertation is the culmination of my hard work, but most importantly, the result of the endless support I received from my mentors, friends, and family. First and foremost, I am forever grateful to my advisor Jakob Jensen; without his constant supervision, mentorship, and encouragement for the last four years and more, this would not be possible. The confidence I have today is largely a result of the empowerment that I received from him. I am also thankful for my exceptional committee members—Kimberly Kaphingst, Ye Sun, Sara Yeo, and Lee Ellington—for their invaluable feedback and guidance throughout the four years. My special gratitude to the Steffensen Cannon Scholarship for funding the final year of my graduate school. The fellowship not only provided financial support but also heightened my self-esteem. Health Communication and Technology (HCAT) lab family is a crucial part of my graduate school. Kevin John’s supervision has been specially instrumental in the lab study with physiological measures. The countless interactions I had with Lisa, Courtney, Mindy, Kate, Andy, and Nick during several writing retreats and conferences have been very helpful in navigating graduate school and the job market. I feel extremely lucky to be a part of this academic family. Throughout the last four years, the Graduate Advisor Amity Mower has always made sure that I was on top of administrative policies, international student rules, and also made sure I submitted all the required forms on time; I appreciate her genuine intentions, regular reminders, and uplifting nature. I am grateful to my graduate friends, who made an invaluable impact. Diana Zulli has a special place in my heart. Our never-ending conversations about graduate school, personal life, culture, and female empowerment have made me a stronger person. I am forever indebted to her for her friendship and continuous motivations. Graduate school would not be the same without Chelsea Ratcliff. For the last five years, she has been my ally, helping me write better, pronounce things correctly, and providing cultural/normative inference to words and dialects. Most importantly, she has listened to my rants and frustrations and motivated me with her kind words. Meaghan Mckasy is an indispensable part of my Ph.D. life. Going through this process together, having her by my side (literally) since day one, and navigating the dissertation and job market together made a significant—actually, statistically significant—difference! Finally, I feel fortunate to have a phenomenal family support system. The love of my life, my husband Jay, is the epitome of love, kindness, and empowerment. His insights, perspectives, and motivations made me a better graduate student. He is my rock and always made sure I did not collapse under the pressure of school and family responsibilities. I am eternally indebted for the role he played throughout graduate school. I cannot imagine doing this without his comfort, supportive gestures, culinary skills, and encouraging words. My parents are the pillar of my strength and the reason for xiii my accomplishments. In a society where daughters were and unfortunately are often perceived as less, they raised a daughter who could dare to dream big and work hard to achieve her dreams. I am forever grateful that they fought with their insecurities, went against societal norms, and believed in their little girl by letting her fly thousands of miles to pursue her dreams. My sisters—Anu and Ranu—are the spark of my life. Despite the distance, they have always found ways to connect with me and keep inspiring. Finally, the support, understanding, and encouragement that I received from my in-laws over the last five years meant a lot. Since my very first day in the Bhandari family, I have felt tremendous love and support. My father-in-law (buba) would have been so proud and happy had he been alive to witness this day. I am sure he is bestowing me with his love and blessings from the heaven. Overall, this dissertation is the result of the immense support and encouragement of numerous people and entities. It did take a village to make the dreams of this little girl true. My sincere acknowledgment to everyone who made this venture possible. xiv CHAPTER 1 INTRODUCTION Visuals have been used as an effective aid in a wide range of fields such as communication, marketing, and information systems (An, 2007; King, 2015a; Thomas, Rompay, De vries, & Van Venrooji, 2010). While visual research is important to numerous domains, this dissertation focuses on visuals as a means of communicating health information, referred to, henceforth, as visual health communication. Visual health communication is defined as an area of study focused on research and practice that utilizes imagery to communicate health-related issues with the intention to create positive impacts on knowledge, attitudes, and behaviors (McWhirter & Hoffman-Goetz, 2014). Existing literature proclaims the potency of visuals in communication interventions (Ancker, Senathirajah, Kukafka, & Starren, 2006) postulating that, among other things, visuals help to draw attention (Smerecnik et al., 2010), summarize complex information (Lipkus & Hollands, 1999), and increase the memorability of messages (Lipkus, 2007). Visuals have promising usage in the field of health communication, including cancer communication (King, 2015b). Within the context of cancer communication, one of the critical uses of visual is to communicate risk information because understanding cancer-related risk involves interpreting complex medical facts based on probabilities (Rothman & Kiviniemi, 1999). To this end, visuals—such as graphs and icon arrays— 2 have the capacity to simplify probability-based complex information (Hildon, Allwood, & Black, 2012). Messages about cancer often include visual information, though corresponding research explicating visual features and effects remains relatively scare (King, 2015b). For example, skin cancer prevention campaigns are increasingly utilizing ultraviolet (UV) photos to convey skin cancer risk. The video “what the sun sees”—which has garnered 81,000 likes to date—depicts individuals viewing and reacting to UV images of their face. Researchers have demonstrated that UV photos are meaningful; Gamble and colleagues (2012) found that skin damage captured by UV photos was related to risk of developing melanoma in 10-12 year olds. However, only a handful of studies have examined how audiences react to UV imagery, and whether these images can be leveraged to shape skin-cancer-related attitudes, beliefs, and behaviors (Gibbons, Gerrard, Lane, Mahler, & Kulik, 2005; Mahler, Kulik, Gerrard, & Gibbons, 2007, 2013). This dissertation launches a research program that aims to explore the impact of visual health communication in the context of cancer. To initiate the program, three studies will be carried out. The first study—reported in the second chapter—is a metaanalysis that explores the efficacy of visual interventions in communicating cancer risk. The goal of the meta-analysis is to quantify the impact of visuals in cancer communication. The second and third studies—reported in the third and fourth chapters, respectively—will focus on understanding the impacts of a specific form of visuals (UV photos), and a specific type of cancer (skin cancer) to evaluate a sound behavior theory (extended parallel process model). The second study is a web-based survey experiment studying the impact of a stock ultraviolet (UV) image in promoting sun-safe behaviors. 3 The third study is a lab experiment that expands on the second study by adding a personalized UV photo condition and evaluating the impact of the imagery using physiological measures (e.g., eye tracking, galvanic skin response). Taken together, these three projects will thread together my interest in visual interventions, theory evaluation, and cancer communication. The goal of this chapter is to set up the research program and explicate the structure of the dissertation chapters to follow. To that end, the next section begins with a brief outline of each study in the dissertation. After that, I will provide a brief review of the research at the heart of the dissertation related to cancer, notably skin cancer. The review will specially focus on cancer risk communication and visual research. Thereafter, an overview of potential theoretical frameworks to guide the program is presented followed by a discussion about the optimal theories to guide the UV photo interventions. Finally, implications of the dissertation and future directions for the research program are outlined. 1.1 Dissertation Chapter Outline The second chapter reports a meta-analysis that synthesizes empirical findings by investigating the effect of using visuals in cancer risk perception. The studies will be screened using three selection criteria. First, it should be studied in the context of cancer. Second, the study should employ an experimental design in that the study compares one or multiple formats of visuals with nonvisual information. Finally, the dependent variable of the study should be related to the perceptions of cancer risks (e.g., risk perceptions, risk estimates, knowledge about cancer risk). Subgroup analysis will also be conducted 4 that will examine the difference of effect size based on the types of visuals and other sample characteristics. The second study—reported in the third chapter—will investigate the effects of stock UV photos at influencing the perception of threat, fear, and expectations related to skin cancer prevention behaviors, and provide an opportunity to test and refine a theoretical framework—the extended parallel process model (EPPM). Thus, a web-based experiment will be conducted to evaluate the influence of stock UV images with a large sample size of about 2000 adults. This study will provide strong statistical power to evaluate the impacts of several message features including UV imagery, and to evaluate and refine EPPM. The study reported in Chapter 4—a lab experiment—expands the second study in two important dimensions. First, an intervention group receiving a personalized UV image will be added, and second, physiological responses (i.e., skin conductance and facial expressions) will be tracked when participants are viewing the visual. The second and third studies inform and complement each other in multiple ways. For example, the control image that is found most influential in the second study will be used as a control stimulus in the third study. Further, the lab experiment will have a smaller sample (about 30-40 participants in each group); therefore, it will have weak statistical power to conduct multivariate analysis. However, this limitation will be offset by the large sample size of the second study. The second study, however, can only utilize self-report measures. Accordingly, the physiological measures incorporated in the third study will advance the findings of the second study by exploring the relation of theoretical constructs with physiological measures (e.g., arousal, facial expressions, and eye tracking). 5 Collectively, these three studies engage multiple methodological and contextual dimensions to tap into the complexity of studying visuals depicting risk information. To that end, one of the complexities associated with studying these types of visuals is how to refer to them. Are these risk visuals because they are communicating risk? For example, the meta-analysis reported in Chapter 2 uses this language/perspective as the synthesized literature primarily comes from risk communication. However, other theories, such as EPPM (Witte, 1992) and HBM (Rosenstock, 1990), frame risk visuals under the umbrella of threat, and that leads some scholars to refer to these visuals as threat appeals (Cauberghe, De Pelsmacker, Janssens, & Dens, 2009; Dillard, Li, & Huang, 2016; Meczkowski, Dillard, & Shen, 2016), but the most common language that scholars use is fear appeals (Tannenbaum et al., 2015; Witte & Allen, 2000), which is how the visuals are referred to in Chapters 3 and 4. To note, these names suggest how people think these visuals function. However, the same visuals are being referenced in multiple ways. This raises a number of questions not only in regard to nomenclature of the visuals but also about the relation of risk visuals to fear appeals. Does risk perception always lead to fear? Or is it the other way around? Can fear exist without the perception of risk? If so, can such fear without the presence of risk trigger behavior change? Indeed, the dissertation program wrestles with these confusions. The metaanalysis studies the impacts of visuals on risk perceptions, whereas the web-based experiment study and the lab study investigates the relation between threat (i.e., susceptibility and severity) and fear. At present, these lexical challenges remain unresolved. This semantic and relational quagmire will likely remain until researchers gain a deeper understanding of the relationship among fear, risk, and threat. It will take 6 considerable time and research to resolve this confusion. The current dissertation is a step in that direction. All in all, the studies of this dissertation are designed to advance the understanding of fear appeal theory, visual message features, and their effects in the area of cancer risk communication. 1.2 Cancer Cancer is a disease condition caused by abnormal growth of body cells. It is one of the major global public health challenges and is the second leading cause of death in the United States (Heron & Anderson, 2015). Some of the most common cancers in United States are breast cancer, lung and bronchus cancer, prostate cancer, and melanoma of the skin (National Cancer Institute, 2017). Approximately 1,685,210 new cases of cancer are estimated to be diagnosed in the United States in 2016 and worldwide, this number is expected to rise to 22 million in the next 2 decades (National Cancer Institute, 2017). The fight against cancer needs ongoing research to improve screening practices, detection technology, treatment, and the promotion of preventive behaviors (Siegel, Miller, & Jemal, 2017). 1.2.1 Cancer Risk Communication Cancer-related mortalities could be prevented by following recommendations such as performing regular screenings and following preventive behaviors (Siegel et al., 2017). However, many people do not follow cancer recommendations, perhaps because they are frustrated, overloaded, and confused (Jensen, Carcioppolo, et al., 2014). In addition, people may ignore recommendations because they perceive themselves to have 7 low susceptibility to a type of cancer. In other words, if people perceive that they are not at risk of getting a type of cancer, then they are less likely to follow the recommended preventive behaviors (Ancker et al., 2006; Bandura, 2004). However, making cancer threats salient is challenging, as it involves interpreting complex medical uncertainties and probability-based risk information (Rothman & Kiviniemi, 1999). In addition, past research has found that people tend to rely on their gut feelings when they need to make decisions revolving uncertainties (Gigerenzer, 2007). Making sense of uncertainties is also dependent on several personal factors such as health literacy, education status, and other cognitive and affective factors. Furthermore, existing research suggests that ambiguous information might lead to confusion, skepticism, and risk-aversion among low literate pessimistic individuals (Han et al., 2011; Politi, Han, & Col, 2007). However, other research has demonstrated that properly communicated uncertainty can bolster the perceived credibility of the communicator and reduce overload and backlash for the audience (Jensen, 2008; Jensen et al., 2011; Jensen & Hurley, 2010). In short, communicating cancer risk is central to cancer control efforts, but it is neither straightforward nor simple to effectively convey risk to target groups. 1.3 Visual Research In health communication, visuals have been used in a variety of contexts, such as the influence of photographic images of obese persons on attitude (McClure, Puhl, & Heuer, 2011), the influence of visual image dose (frequency of images) on skin selfexamination behaviors (King, Carcioppolo, Grossman, John, & Jensen, 2015), cancer risk communication through maps depicting risk levels in different geographical areas 8 (Severtson & Myers, 2013), and effects of message formats in risk perception of alcoholrelated cancers (Chen & Yang, 2015). Visuals are helpful to draw people’s attention (Smerecnik et al., 2010), summarize information efficiently, make hidden facts salient (Lipkus & Hollands, 1999), and especially beneficial when targeted to participants with low literacy (Peters, Hibbard, Slovic, & Dieckmann, 2007). Further, visuals are associated with the memorability of the message, helping with cognition, evoking emotions, and/or persuading behavior change (Lipkus, 2007; Messaris, 2003). Visuals have also been used as an aid to underscore key ideas and reduce distractions (Houts, Doak, Doak, & Loscalzo, 2006). In sum, visuals have promising usage in the field of health communication. Thus, quantifying the effectiveness of using visuals in communicating cancer risk through meta-analysis is a crucial step, which will be accomplished in the first study. 1.4 Skin Cancer Prevention Visual research can be studied in numerous contexts; however, the second and third studies of this dissertation focus on the communication of skin cancer risk. Studying skin cancer is important because its incidence is soaring in the United States, making it one of the most common type of cancer (National Cancer Institute, 2016). Exposure to UV radiation is associated with the three most common types of skin cancer: basal cell carcinoma, squamous cell carcinoma, and melanoma (Gandini, Sera, Cattaruzza, Pasquini, Picconi, et al., 2005). Despite the risk factor of sun exposure and indoor tanning being highly preventable, the cases of the most dangerous type of skin cancer— melanoma—is projected to increase over the next 15 years (Guy et al., 2015). By 2017, 9 the expected deaths linked with melanoma is estimated to be 13,590 (Siegel et al., 2017). Therefore, CDC has called for evidence-based interventions to promote behaviors that reduce UV radiation exposure (Guy et al., 2015). Health promotion campaigns designed to mitigate the mortality rates related to skin cancer are focused on reducing UV exposure or early diagnosis of skin cancer. To reduce people’s exposure to UV rays, these preventative behaviors are advocated: wearing sunscreen, using protective clothing, avoiding the sun, and not using tanning beds (National Cancer Institute, 2016). For early diagnosis of skin cancer, the messaging is focused on underscoring the importance of routine skin self-examination (SSE) and clinical skin-examination (CSE). To note, these campaigns are attempting to address the personal and familial risk factors (e.g., family history, skin type, race/ethnicity) that cannot be prevented by promoting sun-safe behaviors and screening behaviors (Gandini, Sera, Cattaruzza, Pasquini, Abeni, et al., 2005; Gandini, Sera, Cattaruzza, Pasquini, Picconi, et al., 2005; Gandini, Sera, Cattaruzza, Pasquini, Zanetti, et al., 2005). 1.4.1 UV Photo Intervention An emerging approach that seems promising in skin cancer prevention is the use of UV photos that reveal skin damage as eye-catching dark spots. UV photography can provide a personalized visual that helps in public health campaigns and patient education and counseling (Gibbons et al., 2005; Mahler et al., 2007, 2013). One of the challenges of skin cancer prevention is that UV exposure, and skin damage, occur slowly over a long period of time. In fact, some people purposefully expose themselves to high levels of UV via tanning beds because they think it enhances their appearance (Beasley & Kittel, 1997; 10 Turrisi, Hillhouse, & Gebert, 1998). UV photos have the ability to make even immediate wrinkles and dark spots observable, which counteracts the perception of short-term appearance enhancement, and potentially conveys the possibility of long term damage, which might change perceptions (Hollands, Hankins, & Marteau, 2010). Moreover, the dark spots seen in UV photographs are related to the phenotypic risk of having melanoma; individuals showing skin damage in the UV photos are at high risk (Gamble et al., 2012). Technological innovations—such as UV photos—are sometimes hindered by cost concerns, but UV photographs can be collected easily and at little expense (Fabrizi, Pagliarello, & Massi, 2008). Thus, UV photography might be a plausible persuasive strategy in the prevention of skin cancer. 1.5 Theory Overview and Selection Researchers who study visual communication have several theoretical frameworks that articulate the impact of visuals, such as exemplification theory and excitation-transfer theory (Zillmann, 1971, 1999). However, existing theories are designed to explicate particular phenomena (e.g., news coverage, horror films) rather than provide researchers with a dynamic message effects model. One path forward is to think about how visuals communicate information. The answer to that question might suggest other theoretical frameworks for studying the impact of visual communication. For example, visuals can convey threat, which should make perceived susceptibility and severity possible pathways to impact. Perceived susceptibility and severity are core constructs of several behavioral theories, such as the health belief model (Hochbaum, Rosenstock, & Kegels, 1952) and the extended parallel 11 process model (Witte, 1992). Visuals can convey a message more indirectly than nonvisuals, which may result in less motivation for defensive avoidance. Accordingly, that would suggest the utility of frameworks such as psychological reactance theory (Brehm, 1966; Brehm & Brehm, 1981; Dillard & Shen, 2005) or the elaboration likelihood model (Petty & Cacioppo, 1986). Visuals can convey information about people, including what they look like, how they behave, and how commonly they engage in an activity. Indeed, visuals often convey this information, which may explain why most UV photo research to date has been situated in the prototype willingness model (Gerrard, Gibbons, Houlihan, Stock, & Pomery, 2008). In order to appropriately situate the research project under a sound theoretical framework(s), I provide a brief introduction to five potential theories and then, in the section that follows, make a case for a particular theoretical approach. 1.5.1 Potential Theoretical Frameworks 1.5.1.1 Health Belief Model (HBM) The health belief model is one of the early theories and a very prominent theory of social sciences. The HBM includes two major part of psychological perceptions, i.e., perceived threat and behavioral evaluations (Skinner, Tiro, & Champion, 2015). Perceived threat entails perceived susceptibility (perception of being vulnerable to a specific health issue) and perceived severity (perception that the health issue is a serious problem), whereas the behavioral evaluations include costs (barriers) and benefits related with the recommended preventive behavior (Hochbaum et al., 1952; Rosenstock, 1990). HBM postulates that when combinations of threat and behavioral evaluations are able to 12 yield positive belief about the recommended behavior, cues to action will finally prompt the execution of recommended behavior (Hochbaum et al., 1952). A review that evaluated the effectiveness of HBM concluded that the theory has not meaningfully added in the success of interventions and this could be because more than half of the studies did not utilize the theory as a whole (Jones, Smith, & Llewellyn, 2014). 1.5.1.2 Extended Parallel Process Model (EPPM) EPPM is considered as a primary theory of fear appeal (Maloney, Lapinski, & Witte, 2011). The central constructs of EPPM are perceived threats and efficacy beliefs (Witte, 1992; Witte, Cameron, Mckeon, & Berkowitz, 1996). Perceived threats are operationalized in the same ways as in HBM (i.e., includes perceived susceptibility and severity). Efficacy belief is comprised of self-efficacy and response efficacy. To this end, self-efficacy is the perceived self-ability to perform the recommended behaviors and response efficacy as the belief that the recommended behaviors are effective in preventing the threats posed. The pathways to behavior change are defined by the interaction of perceived threat and efficacy beliefs. It is postulated that high perception of severity and susceptibility together with high efficacy beliefs will trigger a danger control process, which leads to following recommended behavior (Witte et al., 1996). However, in the absence of these components, a fear control process is activated that results in the rejection of message. It is postulated that interactions of high threat and low efficacy activate fear control process and produce message resistance (Witte & Allen, 2000). 13 1.5.1.3 Psychological Reactance Theory (PRT) Psychological reactance is defined as “the motivational state that is hypothesized to occur when a freedom is eliminated or threatened with elimination” (Brehm & Brehm, 1981, p. 37). The assumption behind the psychological reactance theory is that individuals believe that they have freedom to behave in anyways they want. When they perceive that their behavioral freedom is being endangered, they are motivated to regain their freedom (Brehm, 1966). In an attempt to regain the freedom, individuals tend to increase the thought, feeling, or action that counteracts the recommendations (Quick et al., 2013). Even though the creator of the theory believed that reactance cannot be measured, Dillard and Shen (2005) operationalized the constructs of psychological reactance theory by developing self-reported measures. Thereafter, psychological reactance theory has been the major theoretical framework to study the defensive reactions of any persuasive efforts. 1.5.1.4 Elaboration Likelihood Model (ELM) Elaboration likelihood model is one of the primary theoretical models in information processing. ELM postulates that persuasion occurs through different channels based on whether people are actively thinking about the subject or not (Petty & Cacioppo, 1986). Specifically, it hypothesizes two routes of persuasion: central and peripheral. Within these routes of persuasion, several pathways are operationalized based on the capability to process (knowledge, distractions), nature of processing (argument quality, initial attitude), and dynamics of cognitive structure (thought accessibility and confidence; Petty & Cacioppo, 1986; Slater, 2002). Pathways between central and 14 peripheral routes interact, suggesting that there might be the transfer in the routes in any phase of processing, but the relationships of these processes have not been explicated (O’Keefe, 2013). 1.5.1.5 Prototype Willingness Model (PWM) A central theoretical framework underlying much of the UV photo intervention study is the prototype willingness model (PWM). The model postulates that willingness to perform any behavior is dependent upon the prototype (Gerrard et al., 2008). Prototypes – also referred as social images– are the typical individuals who are perceived by the target population as someone who follow the recommended behavior. That is, if they have positive image about the prototypes, people tend to be more willing to engage in the behavior. Additionally, this pathway from prototype to behavior engagement might occur even when people are not consciously planning to perform the behavior but what is required is the chance to engage in the recommended behavior (Gerrard et al., 2008; Gibbons et al., 2005). Attitudes, subjective norms, and risk prototypes are the constructs of the theory; all of these interact to cause behavioral willingness, intentions, and actions. 1.5.2 Theory to Guide UV Research Two studies in the dissertation examine the effect of UV photos on skin-cancerrelated behavior expectations. Based on a review of several plausible theoretical frameworks, and consideration of the potential benefits of UV photography to skin cancer prevention, it seems logical to conceptualize the intervention as a fear appeal that potentially motivates change in the receiver by increasing perceived threat and fear. 15 O’Keefe (2003) argued that researchers should define message manipulations based on their intrinsic features rather than their effects. In this case, UV photo interventions communicate invisible damage a feature that should equate to increased susceptibility and fear. Accordingly, the EPPM is a good theoretical framework to guide this research because of its clear identification of message features. 1.6 Implications in Communication Literature 1.6.1 Advance Understanding of EPPM This dissertation endeavors to evaluate one of the most influential fear appeal theories: the extended parallel process model (EPPM; Witte, 1992). It further extends understanding of the EPPM by investigating the position of fear (So, 2013). Investigating fear is important as it intersects with many other streams of research such as emotion, information processing, and advertising. In fear appeal studies, it is often the case that manipulating threat is assumed to initiate fear, without actually evaluating it (Thrasher et al., 2016). Even when fear is measured, it is used as a manipulation check rather than a mediator (Tannenbaum et al., 2015). Also, researchers typically rely on self-reported fear measures (e.g., Dillard, Li, & Huang, 2016). Our project will contribute in the literature by measuring fear and relating it with physiological arousal measures (using GSR), and accordingly investigate the position of fear and physiological arousal in the theoretical framework of the EPPM. 16 1.6.2 Incorporates Physiological Measures in Existing Theoretical Framework This dissertation taps into a larger area of communication research by using physiological measures. The majority of communication studies rely heavily on selfreported measures, but it is possible that self-reports might be flawed because of selfdesirability bias and inability of humans to consistently express their thought processes accurately (Dijksterhuis, 2004; Fazio & Olson, 2003). At the very best, self-reported data accounts for conscious awareness whereas physiological data can yield deep information about the unconscious mind and actions (Cacioppo, Tassinary, & Berntson, 2007; Lim & Reeves, 2009). Physiological measures complement and extend the knowledge gained through self-report surveys. Accordingly, the use of physiological measures in communication studies is evolving. However, majority of the studies that incorporate physiological measures either are atheoretical (e.g., Leckner, 2012, Potter & Choi, 2006), mention theories but do not evaluate the theoretical framework (e.g., Sparks & Lang, 2010), or have used physiological constructs as dependent variables (e.g., Bailey, 2015). In the third study of this dissertation, I attempt to physiologically measure arousal using galvanic skin response (i.e., skin conductance), and emotional valence using facial expression analysis. These physiological variables will then be evaluated in the theoretical framework of the EPPM as potential mediators. 1.6.3 Advance Understanding of Visual Message Features The first study of this dissertation assesses the significance of visuals in communicating risk information through meta-analysis and the study reported in third 17 chapter studies the effects of several types of visual messages (i.e., images of mole excision, sun burn, photo-ageing, sun exposure, and stock UV images) in promoting sunsafe behavior expectations. Thus, these studies will help us understand the effects of each of these message features including the significance of including visuals in risk communication. 1.6.4 Investigates the Utility of UV Photographs in Promoting Sun-safe Behaviors Previous studies have used personalized UV images and demonstrated its effectiveness in communicating risks associated with skin cancer (Gibbons et al., 2005; Mahler et al., 2007). This dissertation will extend this finding in several important ways. First, stock UV photos have not been used so far in research and hence I do not know the significance of providing personalized UV photos versus any other stock UV photos. The study in Chapter 3 of this dissertation compares the effectiveness of stock UV image with other types of visuals used in skin cancer prevention sites and then the study in Chapter 4 compares the effectiveness of personalized UV image with stock UV image and a nonUV visual. 1.6.5 Future Implications in Information Processing Literature There might not be any direct implications of this project in the realm of information processing, but the findings of this study will be valuable to the information processing literature in the long run. Particularly, the findings of the dissertation might synthesize the message processing of fear appeal stimuli. Studies that have attempted to 18 understand processing of fear appeal messages have found mixed findings. Some studies have found that strong threat inducing messages activate defensive systemic or central processing (Liberman & Chaiken, 1992) and others have reported that fear appeal messages provoke peripheral or heuristic processing (Hale, Lemieux, & Mongeau, 1995). Witte and Allen (2000) encouraged future research to study information processing of fear eliciting messages. Thus, the insights that we gain from this dissertation about fear appeal message, defensive motivations, and emotion including physiological arousal might help to design a research program to explore the information processing pathways of fear inducing message incorporating important bio-metric constructs such as physiological arousal and reactance. 1.7 Conclusion Visual health communication is a compelling area of research as it focuses on visuals and explores how they impact people’s cognition, attitude, and behavior. The current dissertation advances our understanding of visual health communication via three studies that focus on assessing the influence of visual communication of cancer threats. The primary research aim of the dissertation is to explicate persuasive impact of visual message features on knowledge, attitudes, and behavior expectation through perceived fear, physiological arousal, and emotional valence. These topics will be studied within the context of cancer communication, as cancer communication is one of the critical areas of health communication. This research program investigates visual message features using a behavior theory (Extended parallel process model) in the context of cancer risk 19 communication. Thus, this dissertation as a whole cultivates my knowledge in the field of visual research, fear appeal, and cancer communication. CHAPTER 2 THE IMPACT OF VISUALS ON CANCER RISK PERCEPTIONS: A META-ANALYSIS Visuals are often used to communicate cancer risk information (Lipkus, 2007). Communicating cancer risk is a complicated process, as it involves interpreting complex medical facts and probability-based risk information (Fischhoff, 1999). Thus, multiple experiments have examined whether visuals alter cancer risk perceptions compared to text. The current meta-analysis synthesizes that research to identify key findings and future directions. 2.1 Visual Health Communication Visual health communication is an area of research and practice that utilizes visual imagery to communicate health information and create positive impacts on knowledge, attitudes, and behaviors related to health (McWhirter & Hoffman-Goetz, 2014). A well-designed visual can grab attention (Smerecnik et al., 2010), present latent facts (Lipkus & Hollands, 1999), and reach low numerate individuals (Peters et al., 2007). Visuals have been used as an effective aid in a wide range of contexts not only in health communication but also in fields such as marketing, psychology, and information 21 systems (An, 2007; Thomas et al., 2010). Scholars have studied visuals in a variety of contexts including the influence of photographic images of obese persons on attitude (McClure et al., 2011), influence of visual image dose (frequency of images used) in skin self-examination behaviors (King et al., 2015), cancer risk communication through maps depicting risk levels in different geographical areas (Severtson & Myers, 2013), and table and bar graphs to communicate health risks (Chen & Yang, 2015). Visuals create ocular stimulation (Lester, 2013), so they are helpful to grasp viewers’ attention. Visuals also enhance message memorability, create cognitive analogies, help weigh the pros and cons, trigger emotions, and/or induce behavior change (Lipkus, 2007; Messaris, 2003). Health educators are advised to use visuals in their interventions as an aid to substantiate their key idea, reduce explanations that are distracting, and include captions in conjunction with pictures (Houts et al., 2006). However, it is challenging to develop visuals that grab attention, provide accurate information, and communicate possibilities and uncertainties (Spiegelhalter, Pearson, & Short, 2011). Moreover, some negative outcomes are associated with the use of visuals— for example, visuals that make fatal consequences more salient and consequently trigger risk aversion (Visschers, Meertens, Passchier, & Vries, 2009). Thus, it is crucial to identify different mechanisms and methodological components of visual impact (McWhirter & Hoffman-Goetz, 2014). Researchers have expressed the need for a systematic exploration of visual messages and their effects (Jensen, 2011; Yzer, 2011). One way of conducting a systematic investigation is to analyze the average effect size across all the studies comparing visuals with texts. Indeed, health educators are advised to evaluate the 22 effectiveness of messages with and without pictorial stimulations (Houts et al., 2006). There are several narrative reviews that summarize the findings in visual health communication (Ancker et al., 2006; Houts et al., 2006; McWhirter & Hoffman-Goetz, 2014), and this meta-analysis complements that work by providing a quantitative synthesis (Allen, 2009; Schmidt & Hunter, 2014). To date, narrative reviews support the idea that, compared with text alone, visuals increase accurate cognitions of cancer risk. H1: Compared with text alone, text plus visuals increase perceptions of cancer risk. 2.2 Moderators 2.2.1 Visual Format King (2015a) advocated for more research studying the impact of visual format such as icon arrays, bar/pie charts, and photos. A systematic review of 30 studies concluded that icon arrays, bar/pie charts, and tables are the most common formats used in visual research and icon arrows are the most preferred format and also better comprehended by participants (Hildon et al., 2012). Other studies, however, found that both icon arrays and bar charts yield comparable risk estimates and both outperformed other forms of visuals (Garcia-Retamero & Galesic, 2010). Of note, Hawley and colleagues (2008) reported that all forms of visuals were rated positively by participants, yet icon arrays were perceived as more trustworthy by individuals with low and high numeracy. In another study, participants preferred pictorial depiction when breast cancer risk was presented alone, but bar charts were preferred when breast cancer risk was presented in comparison to heart disease, stroke, and osteoporosis (Schapira, Nattinger, & McAuliffe, 2006). A review article by Ancker and colleagues (2006) recommended that 23 future research should identify visuals that are both preferred and effective. 2.2.2 Sample Characteristics Next, it is important to consider the impact of visuals based on differences in sample characteristics. Past research has found that females tend to perceive higher risks than males (Finucane, Slovic, Mertz, Flynn, & Satterfield, 2000). Young adults are perceived as risk averse compared with older adults (Green, Mitchell, & Bunton, 2000). Also, health literacy and numeracy, which impact comprehension of risk information, are lower among elderly populations (Baker, 2007). 2.2.3 Sampling Method Furthermore, the use of web-based professional recruitment systems (e.g., Mechanical Turk, Qualtrics Panels) is increasingly common to recruit samples for studies. These systems provide researchers with the opportunity to recruit large and diverse samples as compared with face-to-face recruitment. However, web-based panels are also susceptible to manipulation and potentially bias (Couper, 2000). All of which raises questions about study differences based on the sampling method, which brings to the first research question (RQ). RQ1: Compared to text alone, does the impact of text plus visuals vary according to visual format (RQ1a), sample characteristics (RQ1b), or sampling method (RQ1c)? 24 2.3 Method 2.3.1 Article Identification and Selection A systematic search of related articles was conducted in the following databases: Communication and Mass Media Complete, PsycINFO, MEDLINE, PsycArticles and CINAHL using this search phrase (Visual* OR Pictograph* OR iconarray OR "Icon array" OR illustration* OR graph* OR chart* OR infograph*) AND (Experiment*) AND (knowledge OR "risk perception*" OR “risk estimate*”) OR "risk literacy" AND (cancer*) for all studies published on or before August 2016. The search located 210 items. After the initial screening, 20 were duplicates, and 159 items were found to be irrelevant. There was a book in the search list, which was excluded from the pool. The remaining 28 papers and two dissertations were screened using three selection criteria. First, it should be studied in the context of cancer. Second, the study should employ an experimental design in that the study compares one or multiple formats of visuals with nonvisual information because the primary interest is in examining the experimental effects of using visuals versus nonvisuals. Finally, the dependent variable of the study should be related to the cognitive aspect of cancer risks. Risk perceptions, risk estimates, or knowledge about cancer risk are some of the examples of the cognitive aspects that could be used as the dependent variable. After a careful screening based on the above three criteria, a final sample of six articles were retained (Chen & Yang, 2015; Han et al., 2012, 2011; Pepper, Cameron, Reiter, McRee, & Brewer, 2013; Waters, Weinstein, Colditz, & Emmons, 2007; Zikmund-Fisher et al., 2008) as 21 articles and two dissertations were not included as they did not meet the dependent variable and/or study design criteria. I reached out to the corresponding 25 authors if we needed additional information. The screening procedure is presented in Figure 2.1. 2.3.2 Unit of Analysis For the analysis of this paper, six articles were reviewed. In the experimental study of the visuals, multiple comparison groups are often reported, so I used each visual versus nonvisual component of the study that captured different conceptual distinctions as a unit of analysis. Using these guidelines for defining effect sizes, the final effect size was eight. 2.3.3 Coding To examine the conditions under which the influence of visuals differentiates, three sets of moderators were established: visual format, sample characteristics, and sampling method. The development of the codebook was guided by these principles. First, I wanted to explore the factors that would contribute theoretically to the visual communication of risk literature. For example, identifying the most beneficial form of visual stimuli across all the studies. Secondly, I examine the moderating role of population characteristics such as sex and age of samples. Finally, I wanted to explore the features of research procedures that may impact effect sizes (i.e., recruitment methods and other characteristics of research methods; Schmidt & Hunter, 2014). The codebook is presented in Table 2.1. It includes the variables, their definitions, categories, and values assigned with the instructions provided to the coders. These moderators were coded for each effect size. I conducted five subgroup analyses for each 26 of the five variables coded in the study. I am interested in assessing whether the effects of visuals differentiate based on other factors such as sample characteristics (e.g., sex, age of the participants) and sample recruitment (i.e., web-based versus in-person recruitment). I also studied the difference between using two different types of visuals (i.e., icon arrays and graphs). All the moderator variables are presented in Table 2.2 with their operational definitions and intercoder reliabilities. Two independent coders (including myself) coded these five variables from six randomly selected articles and achieved Krippendorff’s alpha of .82 to 1.00, which exceeds the benchmark for reliability (which is .67; Hayes & Krippendorff, 2007; Krippendorff, 2004). The coders resolved discrepancies in the coding through discussion, and the codebook was revised accordingly. Then, I completed the coding for the remaining articles. 2.3.4 Extracting Effect Sizes Each visual versus nonvisual component of the research was coded as an individual effect size. For example, if multiple visuals were used in the same study to compare their effectiveness against the text, then the comparison of each type of visual against the nonvisual (e.g., texts) was used as an individual effect size. The types of visuals I was interested in were icon arrays and graphs. Three kinds of statistical information from the articles were obtained: means and standard deviations of the experimental and control group, odds ratio, and Pearson’s correlation r. All the statistical tests were calculated in a common metric d. For the studies with continuous data that provided means and standard deviations of the visual 27 and text groups, I calculated d using a formula from Borenstein and colleagues (2009). In one of the studies, visual and textual groups were divided into two groups based on the presence and absence of ambiguity in experiment 1 and two types of visual (Han et al., 2011). In this case, I used pooled mean and pooled standard deviation cases to calculate the overall means and standard deviations of the visual and text groups (Borenstein et al., 2009). Additional data were requested by contacting authors where sufficient information was not included in the primary publications. The final effect sizes, variance, and standard error for all the studies can be found in Table 2.3. 2.4 Results I utilized six articles (eight effect sizes) with 5,382 total participants. The effect sizes ranged from -.10 to .28. Effect size distribution is presented visually in Figure 2.2. 2.4.1 Overall Effect Size I report the findings using a random effects model. In a random effects model, the true effect is allowed to vary amongst studies because study samples and interventions will vary and this will have an effect (Borenstein et al., 2009). The standardized mean difference (Cohen’s d) is .11, Z = 3.97, p < .001 (twotailed), with a 95% confidence interval ranging from .05 to .16. The effect size is characterized as small (Cohen, 1988), but statistically different than zero. Thus, H1 is supported: compared to text alone, text plus visuals increase perceptions of cancer risk. 28 2.4.2 Test for Heterogeneity In order to test for heterogeneity, I calculated the weighted sum of squares. The weighted sum of squares was not significant, Q(7) = 6.33, p = .50. The between-studies variance is measured by T2, which is equal to .001 (SE = .004). The 95% confidence interval for T2 ranges from -.008 to .515. The between-studies standard deviation is T = .03, and 95% of the true effects will fall in a range of the confidence interval from .09 to .72. The I2 value is -10.51%, which represents the proportion of observed dispersion because of the heterogeneity. Also, the 95% prediction interval for true standardized mean in a future study is -.08 to .13. In summary, as the p-value for Q is not significant, T2 is low, and I2 is low, there is limited evidence that effect sizes vary across studies. Given the lack of heterogeneity, there is insufficient justification to support moderator analysis (Cortina, 2003). Thus, in response to RQ1a – RQ1c, the impact of text plus visuals does not vary according to visual format, sample characteristics, or sampling method. 2.5 Discussion This meta-analysis reviewed experiments with visual versus textual representation of cancer risk in the context of cancer risk. The findings support the first hypothesis of the study that using visuals increases perceptions of cancer risk. However, I could not proceed to subgroup analysis as the test of heterogeneity failed. These findings underscore the value of visuals in cancer risk communication. Visuals have been considered a key component of message design in risk communication; the results of the current meta-analysis support that adding visuals to a 29 risk message will provide a small, but consistent, increase in perceived risk. Moderator analysis was not viable in the current analysis, but researchers should continue to explore the possibility that the impact of visuals on risk perception is contingent on other factors. For instance, scholars of visual communication should explore the impact of visual dose (or frequency), quality, and density as possible moderators (King, 2015b). Researchers should also be mindful of situations where visuals might have a negative impact on risk perceptions. For example, Fagerlin and colleagues (2007) caution that not all visuals yield positive findings. Indeed, the identification of visual features or message contexts that invoke negative audience response is paramount to effective risk communication. One possibility that warrants additional exploration is the relationship between affect and risk visuals. Do individuals with high trait anger respond differently to risk visuals? Past research has shown that individuals with high trait anger are more likely to interpret uncertain or ambiguous situations as a threat, which raises the possibility of a negative response (Dillon, Allan, Cougle, & Fincham, 2015; Wilkowski & Robinson, 2010). This study is limited in number of ways. First, the total number of studies for this synthesis is not large; this significantly drops the power of the statistical tests. Second, only published articles were used. Grey area literature was not explored. This could be addressed by future meta-analysis by broadening the horizon of the inclusion criteria. Similarly, including all studies focused on health-related risk – not just limited to those in the context of cancer — would help to yield more effect sizes that produce generalizable findings. Third, I did not differentiate communication of numerical risk and narrative risk, which might be an important distinction. Finally, I did not explore the results based 30 on literacy level as visuals are believed to be helpful especially to people with low literacy levels (Houts et al., 2006). Visual representations of numerical information is recommended to address low literate population (Garcia-Retamero & Galesic, 2010; Peters, 2012). As more than half of the Americans have intermediate health literacy levels (Kutner, Greenburg, Jin, & Paulsen, 2006), exploring the effects of visuals for communicating numerical risk in low literate population through meta-analysis would be helpful. 2.6 Conclusion The results of this meta-analysis support the use of visuals in cancer risk communication. Researchers and practitioners should assume that visuals will increase accurate perception of cancer risk. Future research exploring moderators and situations where visuals do not increase perceived risk would be beneficial to the field. 31 Table 2.1 Codebook CODING INSTRUCTIONS Variables Definitions Journal Name Paper ID Paper identification code Study ID Study identification code Effect ID Effect size identification Year Year of publication Age Age range of participants Recruitment Visuals Where are the patients recruited from? The type of visual stimuli used in the study. Categories Values Instructions Each visual vs. nonvisual effects Young adults Older adults General population Web-based panel 1 Direct recruitment 2 Graphs 1 Photo 2 Icon array 3 Mixed 4 2 3 1 75% of participants are younger than 40 years 75% of participants are older than 40 years Mixed population of adults all age range. Participants are recruited through a web-based survey panel like Mechanical Turk. The participants are recruited through and institutions like hospital or university. Bar or pie chart is used as a visual component of the stimuli A photo or picture is used as a visual component of the stimuli Icon array (use of dots, symbols, or picture to depict probability, percentage) or pictograph (i.e., pictorial representation of statistics) is used as a visual component of the stimuli More than one kind of visual is used in the same message group. 32 Table 2.2 Variables, Operational Definitions, and Intercoder Reliability Variables Sex Definitions The sex of the participants. Categories Males α .82 Females Mixed Age Age range of participants. Young adults 1.00 Older adults General population Recruitment Where were the patients recruited from? Web-based 1.00 Direct recruitment Visuals The type of visual stimuli used in the study. Icon-array 1.00 Graph Photo Note. Intercoder reliability was calculated for each variable using Krippendorff’s alpha. 33 Table 2.3 Study Effect Sizes, Variance, Standard Error Study ID Study d Vd SEd 1 (Pepper et al., 2012) .00 .02 .14 2 (Waters et al., 2007)_Chart .07 .00 .05 3 (Waters et al., 2007)_Icon array .16 .00 .05 4 (Chen & Yang, 2015) .28 .03 .18 5 (Han et al., 2011)_Exp-1 -.05 .02 .13 6 (Han et al., 2011)_Exp-2 -.12 .03 .18 7 (Han et al., 2012) .13 .02 .15 8 (Zikmund-Fisher et al., 2008) .14 .01 .08 Note. We calculated d using a formula from Borenstein and colleagues (2009). 34 Search results from September 2016 (N = 210) After initial screening of title and abstract, Irrelevant = 159 articles and a book Removed 18 articles and 2 dissertations as message manipulation was not visual vs. text Removed 4 articles, as dependent variable was irrelevant Screening based on all three screening criteria removed 22 articles and 2 dissertations. (N = 6) Figure 2.1 Screening flowchart 35 8 7 6 5 4 3 2 1 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 Figure 2.2 Effect size distribution 0.2 0.25 0.3 CHAPTER 3 DO ULTRAVIOLET PHOTOS INCREASE SUNSAFE BEHAVIOR EXPECTATIONS VIA FEAR? A RANDOMIZED CONTROLLED TRIAL IN A SAMPLE OF U.S. ADULTS 1 3.1 Introduction Visuals are often used to communicate skin cancer risk because aspects of the risk, and the cancer, are visible (i.e., moles, mole removal scars) or appearance based (i.e., tanning behavior). This explains the abundance of visual images in skin cancer prevention campaigns and social media as well as the pursuit of innovative approaches to visualizing skin cancer risk. Concerning the latter, ultraviolet (UV) photos (i.e., photographs taken using UV light) reveal skin damage as dark spots and patches, which are caused by the agglomeration of melanin just beneath the skin’s surface. These UV photos can be used to assess and communicate melanoma risk (Gamble et al., 2012; Hornung & Strecher, 2012). Accordingly, researchers have utilized UV photos in skin cancer prevention interventions. A meta-analysis found that UV photo interventions increased sun-safe behavior and perceived susceptibility to aging of the skin (also known 1 Reprinted by permission from Springer Nature: Springer Journal of Behavioral Medicine (Do ultraviolet photos increase sun-safe behavior expectations via fear? A randomized controlled trial in a sample of U.S. adults, Pokharel, M., Christy, K. R., Jensen, J. D., Giorgi, E., John, K. K., & Wu, Y. P), (2019). 37 as photoaging; Williams, Grogan, Clark-Carter, & Buckley, 2013). The meta-analysis did not examine fear, as no available study had assessed it in an experimental design. UV photos could be conceptualized as fear appeals, as they can cause viewers to reflect on the sun damage present in their own skin. Researchers have devoted considerable resources to studying fear appeals, or attempts to influence behavior by providing frightening information (Shen & Coles, 2015; Witte, 1992, 1994; Witte & Allen, 2000). In pursuit of this goal, researchers have developed and refined numerous fear appeal theories over the past 50 years (Tannenbaum et al., 2015). The extended parallel process model (EPPM) represents a synthesis of earlier fear appeal theories, and has become one of the cornerstone frameworks for the study of fear (Maloney et al., 2011). Different message types have been evaluated using the EPPM, including texts (Krieger & Sarge, 2013), visuals (Jain, Hoffman, Beam, & Xu, 2017), audio (Birmingham et al., 2015), and audiovisuals (Shi & Smith, 2016). Of these, scholars have noted a need for more research exploring visual fear appeals (King, 2015a). The current study contributes to research on UV photo interventions, the EPPM, and visual fear appeals by comparing the persuasive impact of UV photos of skin damage to other visuals that convey skin cancer risk, develops and tests a UV efficacy visual (e.g., a person putting on sunscreen in UV light), and tests fear as a mediator. In a larger sense, the study compares five different types of skin cancer visuals (UV skin damage visuals, visuals of people engaging in sun exposure, visuals showing sunburned skin, visuals depicting aged skin, and mole removal incision visuals) and thus provides a foundation for future research focused on visual communication and skin cancer prevention. 38 3.2 The EPPM: A Framework for Studying Fear Appeals The EPPM postulates that threat, efficacy, and fear all play a role in the processing of fear appeals (Witte, 1994). In the EPPM, threat is comprised of severity and susceptibility, where severity is the perceived magnitude of the negative consequences related to a hazard, and susceptibility is the perception of self-vulnerability to a hazard (Witte, 1994). Efficacy beliefs are comprised of self-efficacy and response efficacy, where self-efficacy is the perception of one’s own ability to perform recommended behaviors, and response efficacy is the belief in the effectiveness of those behaviors in reducing or preventing a threat (Witte, 1994). However, the EPPM is ultimately a framework for studying fear appeals and, fittingly, fear is central to the model. Within the EPPM, fear is defined as “… an internal emotional reaction comprising psychological and physiological dimensions that may be aroused when a serious and personally relevant threat is perceived,’’ (Witte et al., 1996, p. 320). Fear is generated by demonstrating the negative consequences of a particular behavior in an atrisk individual; for example, a common smoking cessation fear appeal is to show imagery of a smoker’s tar coated lung as compared to a nonsmoker’s healthy lung. The EPPM postulates that the interaction of perceived threat and efficacy will ignite a fear response; if threat is sufficiently mitigated by the efficacy portion of the message, it activates a danger control process that leads to the adoption or acceptance of the behaviors promoted by the intervention (Witte, 1992; Witte et al., 1996). However, threat unmitigated by efficacy activates a fear control process that generates uncontrolled fear, which ultimately leads to rejection of the message. A review article summarizing six meta-analyses concluded that self-efficacy, response efficacy, and susceptibility are 39 critical in achieving the intended results of fear appeals, whereas emphasizing the severity of a threat was not as impactful (Ruiter, Kessels, Peters, & Kok, 2014). The EPPM is an important theory that has aided in the growth of fear appeal research (Maloney et al., 2011), but additional evaluation and exploration of the theoretical framework can strengthen it even further. To that end, researchers have pointed out a need for more experimental research evaluating fear-appeal theories (Peters, Ruiter, & Kok, 2013). Notably, debate continues on the mediating role of fear within fear appeal theories (O’Keefe, 2003; Popova, 2012). In their meta-analysis of fear appeal research, Tannenbaum and colleagues (2015) specifically note the lack of studies directly testing fear as a mediator: Although many fear appeal theories discuss fear, empirical studies typically test the impact of fear appeal messages on outcomes, and subsequently infer that message effects were mediated by experienced fear even though fear itself is rarely measured (for a discussion, see Popova, 2012, p. 466). Indeed, only 71 of the 248 studies in the current meta-analysis measured fear directly, and such measures were typically treated as manipulation checks rather than independent variables or mediators. (p. 1180) Thus, there is a need for more theoretically grounded research that directly assesses fear as a mediator (not a manipulation check) to determine how it functions in fear appeal research (O’Keefe, 2003; Popova, 2012; Tannenbaum et al., 2015). 3.2.1 Skin Cancer Prevention, Ultraviolet (UV) Photo Interventions, and the EPPM Skin cancer prevention is a suitable context for evaluating the EPPM as there is a heavy reliance on fear-based messaging and visuals (see, e.g., Mays & Zhao, 2016). Skin cancer is the most common type of cancer in the United States (Siegel et al., 2017), and 40 efforts to educate the public about both risks and preventative behaviors rely heavily on visual messages, with significant variance in the forms, features, and categories used in these efforts (King, 2015a). Use of visual messages in skin cancer prevention efforts is common because unlike most types of cancer, skin cancer lesions are outwardly visible. These visual messages often include fear appeal elements, such as showing the results of blistering sunburns or the removal of cancerous lesions. However, it is inherently difficult to visualize the underlying cumulative skin damage as a result of excessive sun exposure, tanning, and blistering sunburns. Most skin damage is long-term and cumulative. The inability to see skin damage in visible light might explain the attraction of tanning. Tanners may value short-term gains in appearance while disregarding the invisible damage to the skin that leads to long-term damage and risk. UV photography offers an innovative solution to the challenge of visualizing skin damage. UV photographs are capable of revealing existing skin damage caused by UV light exposure that is normally invisible to the naked eye. Gamble and colleagues (2012) linked the damage depicted in UV photographs with phenotypic risk of developing melanoma (the most dangerous type of skin cancer), suggesting that UV photographs can help identify high-risk individuals. Since then, UV photography interventions have shown significant results in diverse samples, including children aged 11 to 13 (Demierre et al., 2009) and teenagers (Taylor, Westbrook, & Chang, 2016). Moreover, a metaanalysis found that UV photo interventions have positive impacts on sun-safe behavior and susceptibility to photoaging (Williams et al., 2013). UV photos increase perceived susceptibility to photoaging, but do they increase perceived susceptibility to skin cancer and/or fear? One past study found that UV photo 41 interventions increase perceived susceptibility to skin cancer (Emmons et al., 2011; for a review, see McWhirter & Hoffman-Goetz, 2013). Scholars have highlighted the importance of affect—particularly negative emotions—for interventions promoting sunsafe behaviors through appearance-based interventions (Mahler, 2014, 2015, 2018). Two studies (Mahler, 2014; Mahler, 2018) have assessed negative affect following exposure to UV photos, but only Mahler (2018) included a non-UV photo comparison condition. The evidence in hand suggests that UV photos likely generate negative affect compared to non-UV conditions. However, further investigation to parse out the affective component produced by UV photos is needed. Thus, I evaluate these queries and basic postulates of the EPPM in the first set of hypotheses: H1: Compared to all other visual categories combined, a UV skin damage visual condition will generate greater (a) susceptibility, (b) severity, and (c) fear. UV photos could also target self-efficacy and response efficacy by depicting an individual putting on sunscreen (which appears black in UV light). Utilizing this approach, I included four types of efficacy messages within the current study. Participants received either a no efficacy message, a text-only efficacy message, a UV visual-only message (depicting an individual putting on sunscreen in UV light), or a combined visual and text message. I hypothesize that: H2: Compared to the no efficacy and text-only efficacy message conditions, efficacy messages including a UV efficacy visual will generate greater (a) selfefficacy and (b) response efficacy. The existing findings of UV intervention studies (Emmons et al., 2011; Mahler, 2014, 2015, 2018) mirror larger questions within fear appeal literature as past EPPM 42 research has also questioned whether perceived threat triggers fear (or vice versa), whether fear needs to have a more prominent position in the model, and the ideal positioning and role of fear within the EPPM (Dillard et al., 2016; Dillard, Li, Meczkowski, Yang, & Shen, 2016; So, 2013; Witte, 2013). Indeed, a recent metaanalysis noted that past research has rarely tested the mediational role of fear (Tannenbaum et al., 2015). To this end, Mahler (2018) demonstrated that exposure to a UV photo and comparison photos of peers (with lower skin damage than themselves) generated negative emotions, which then mediated the relationship between condition and sun protection intentions. Skin damage manifests over time, and so do sun-safe behaviors. This presents skin cancer researchers with a challenge as quantifying sun-safe behavior requires a measure of achieved or intended action across time. Recently, Armitage and colleagues (2015) found that behavioral expectation—what a person expects to do—is a better predictor of actual behavior than behavioral intention. Given that, the current study created measures of sun-safety expectations to serve as an outcome to assess the emerging questions generated in EPPM literature. The following hypotheses postulate the mediating role of fear and moderating role of efficacy stimuli: H3: Fear will mediate the relationship between exposure to UV skin damage visuals and sun-safety behavioral expectations, such that participants in the UV skin damage visual condition will report greater fear, which will be positively related to behavior expectations regarding sun-safe practices. H4: Exposure to a UV efficacy visual will moderate the indirect path through fear, such that the indirect effect of UV skin damage visuals on sun-safety 43 behavior expectations is larger for individuals receiving the efficacy visual. Several studies of UV photo interventions have shown that UV photos promote skin cancer prevention behaviors (Gibbons et al., 2005; Mahler et al., 2007, 2013; Walsh & Stock, 2012). These previous studies were based on personalized UV photos (i.e., photos of the person in question), but the effects of stock UV photos (i.e., photos of unknown individuals) would benefit from additional research. The value of stock UV photos is that they are easy to generate/acquire and more comparable to most existing skin cancer visuals (which are typically stock images). Understanding the impact of stock UV photos also helps researchers to better understand the impact and value of personalized UV photos by isolating one message feature (UV) from another (personalized). Thus, the current study compares stock UV skin damage photos with four alternative forms of skin cancer risk visuals (visuals depicting sun exposure, sunburn, photoaging, or mole removal). In addition to examining how UV skin damage visuals perform compared to naturally occurring skin cancer risk visuals in general, it is also valuable to study how UV skin damage visuals perform when compared with each of these discrete, naturally occurring categories. For example, past research has found that visuals depicting the aftermath of mole removal can generate significant fear (Mays & Zhao, 2016). To understand the relative persuasive effects of stock UV visuals in comparison with other naturally occurring skin cancer risk visuals, I asked the following research question (RQ): RQ1: Compared to the UV skin damage visual category, is there another naturally occurring skin cancer risk visual type that yields equivalent or greater impact on (a) susceptibility (b) severity, (c) self-efficacy, (d) response efficacy, (f) 44 fear, and (e) behavioral expectation? 3.3 Method 3.3.1 Participants and Procedure Qualtrics Panels recruited 2,219 adults (age range: 18-89, Mage = 43.49, SD = 15.82) from their national panel into an online message experiment. Approximately 45% of the participants were male. The participants filled out a consent form, completed a pretest, viewed one of the experimental conditions, and completed a posttest. The pretest survey included questions about demographics, susceptibility, severity, self-efficacy, response-efficacy, and behavior expectation. The posttest measured fear, susceptibility, severity, self-efficacy, response-efficacy, and behavior expectation. Participants were offered a small financial incentive by Qualtrics Panels to participate in the study. 3.3.2 Study Design Participants’ were randomized to one of 60 conditions in a 5 (visual factor: UV skin damage visuals, sun exposure visuals, sunburn visuals, photoaging visuals, and mole removal visuals) × 3 (replication factor: three examples of each visual condition) × 4 (efficacy factor: no efficacy, text only, visual only, text + visual) between-participants message experiment. The replication factor was nested within the visual factor. 3.3.3 Stimuli Participants’ stock UV skin damage photos were produced in our lab using a VISIA UV camera system. Members of the research team posed for UV skin damage 45 photos. All comparison group visuals were collected from the educational materials, websites, blogs, and social media pages (Facebook and Instagram) of organizations such as the Skin Cancer Foundation, the American Academy of Dermatology, the Centers for Disease Control and Prevention (CDC), and the American Cancer Society. Our research team found that the educational materials and social media pages designed by governmental and nongovernmental organizations generally used four different types of visuals: visuals focusing on sun exposure, including sun bathing with or without sun protective items (i.e., hats, sunglasses, umbrella); visuals showing sunburns, including a body part or face that is severely burnt; visuals that deliver information about photoaging, such as an image showing a single individual with both photo-aged and nonaged skin; and visuals of cancerous mole removals, including both cancerous moles and surgical excisions. Refer to Supplemental Materials 3.6 1-3.6.5 for all the visual condition stimuli. The efficacy condition had four levels: control (no efficacy stimuli), text-only efficacy stimuli, UV visual-only efficacy stimuli, and UV visual + text stimuli. The text stimuli discussed different behaviors that a person can do to prevent skin cancer, and the visual stimuli depicted a person’s face under both natural light and UV light. Half of the person’s face is covered with sunscreen, which is invisible in the natural light photo but appears black in UV photo. The latter demonstrates the potential for sunscreen to block UV rays from the sun, and prevent them from reaching the skin. Thus, the image has the potential to impact self-efficacy (the individual is depicted putting on sunscreen) and response efficacy (the sunscreen is effective at blocking UV). Efficacy stimuli are included in Supplemental Materials 3.6.6 and 3.6.7. 46 3.3.4 Measures 3.3.4.1 Demographics In the pretest, participants provided demographic information, including age, sex, ethnicity, household income, highest level of education, and marital status. Skin-cancerrelated risk was measured using a brief risk assessment tool (BRAT; Glanz et al., 2003). 3.3.4.2 Behavior Expectations Baseline and posttest sun-safety behavior expectations were measured on a sevenpoint scale ranging from 1 (extremely unlikely) to 7 (extremely likely) (Pretest: α = .85, M = 4.86, SD =1.48; Posttest: α = .88, M = 5.23, SD = 1.45). I measured behavior expectations as opposed to behavior intention based on research by Armitage and colleagues (2015), which demonstrated that what a person expects to do is a better predictor of actual behavior than what a person intends to do. The only difference between behavior expectation and behavior intention is that while measuring the former, participants are asked how likely they are to use various sun-safety behaviors as opposed to whether they intend to use those behaviors. The measures were reworded from behavioral measures of photo protection, which were previously developed and validated (Aspinwall, Taber, Kohlmann, Leaf, & Leachman, 2014). The items ask how likely the individuals were to perform these behaviors in the future—“using sunscreen,” “reapply sunscreen after swimming or perspiration,” “wearing protective clothing (long pants and sleeves),” “avoiding peak UVR exposure from 10 AM to 4 PM,” and “stayed in the shade.” I added two additional items to the scale: “wearing broad brimmed hat,” and “wearing sunglasses.” 47 3.3.4.3 Self-efficacy Self-efficacy related to sun-safe behavior was assessed with nine items measured on a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest α = .90, M = 4.90, SD =1.28; Posttest α = .93, M = 5.15, SD = 1.34). The individual items are modified from two previously used scales. Six of the items were used from Witte (2000): “I am able to use sunscreen with at least SPF-15 or higher to prevent skin cancer,” “Using sunscreen with at least SPF-15 or higher to prevent skin cancer is easy for me,” “Reapplying sunscreen every two hours to prevent skin cancer is convenient for me,” “Reapplying sunscreen after swimming or perspiring to prevent skin cancer is easy for me,” “Wearing a hat that provides shade for my face to prevent skin cancer is easy for me,” and “I am able to minimize my exposure to the sun at midday to prevent skin cancer.” The three remaining items were from a different scale (Heckman, Handorf, Darlow, Yaroch, & Raivitch, 2017): “use sunscreen when I am out in the warm sun for more than 15 minutes,” “use sunscreen when none of my friends are using it,” and “use sunscreen even if I don’t like how it feels.” 3.3.4.4 Response Efficacy Response efficacy was assessed using eight items measured on a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest α = .97, M = 5.29, SD =1.33; Posttest α = .97, M = 5.47, SD = 1.39). The individual items were statements such as “My using sunscreen is effective in preventing skin cancer,” “Applying sunscreen with at least a SPF-15 or higher is effective in preventing skin cancer,” “Applying sunscreen to all areas of my body exposed to the sun is effective in preventing skin 48 cancer,” and “Reapplying sunscreen every two hours is effective in preventing skin cancer.” These items are modified from Witte (2000). 3.3.4.4 Susceptibility Susceptibility was measured using three-items and a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest α = .92, M = 4.41, SD =1.58; Posttest α = .94, M = 4.61, SD = 1.59). These items were developed by Witte and colleagues (1996); “I am likely to get skin cancer sometime during my life,” “I am at risk of getting skin cancer sometime during my life,” and “It is possible that I will get skin cancer sometime during my life.” 3.3.4.5 Severity Severity was also measured by the items developed by Witte et al. (1996). These items are “I believe that skin cancer is a severe health problem,” “I believe that skin cancer is a serious threat to my health,” and “I believe that skin cancer is a significant disease.” These items were measured in a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest α = .86, M = 5.42, SD =1.32; Posttest α = .9, M = 5.52, SD = 1.38). 3.3.4.6 Fear A six-item scale developed by Witte (2000) was used to measure perceived fear produced by the study stimuli. The six items are assessed by a seven-point scale ranging from 1 (not at all) to 7 (very much) (α = .96, M = 2.77, SD =1.75) that asks participants 49 to answer how much the message made them feel – “frightened,” “tense,” “nervous,” “anxious,” “uncomfortable,” and “nauseous.” 3.3.5 Randomization Check I first ran a three-way MANOVA to check whether visual conditions (UV skin damage, sunburn, sun exposure, aging, and mole removal), efficacy conditions (no efficacy, text only, visual only, and text + visual) and the interaction of the two were related to demographic variables (age, education, income) and the BRAT index. The multivariate tests were not significant for visual condition, Pillai’s Trace = .01, F(16, 8760) = .72, p =.78, efficacy conditions, Pillai’s Trace = .01, F(12, 6567) = .87, p = .58, or the visual condition × efficacy condition, Pillai’s Trace = .02, F(48, 8760) = 1.10, p = .30. This supports that the demographic variables and BRAT index were successfully randomized across experimental conditions. Next, a 5 (visual conditions) × 3 (replications) × 4 (efficacy conditions) MANOVA with replications as a nested factor examined randomization for all of the EPPM test variables (except for perceived fear) and sun-safe behavior expectations. Fear was only measured in the posttest because it is a state-based measure (i.e., capturing the impact of exposure to the stimuli). The pretest measures helped us identify if any of the main variables (mediators and outcome variables) were significantly related to any of the experimental conditions (i.e., failure of randomization). The multivariate test was not significant for visual condition, Pillai’s Trace = .01, F(24, 8752) = 1.24, p = .20, efficacy condition, Pillai’s Trace = .01, F(18, 6561) = .76, p = .75, or the visual condition × efficacy condition, Pillai’s Trace = .03, F(72, 13140) = .97, p = .55. 50 Results from these two MANOVAs demonstrate that randomization was successful for the study. Thus, I did not use any of the pretest variables and demographics in the analyses for the hypothesis testing. It is important to note that the basic results of the study do not change if the pretest variables are included as controls. 3.3.6 Power Analysis G*Power was used to identify the optimum sample size for the study a priori (Erdfelder, Faul, & Buchner, 1996). G*Power estimated that a sample size of 1,788 was required to detect a small effect size of .10 with power of .95. Our final sample size was 2,219. 3.4 Results 3.4.1 Bivariate Correlations Bivariate correlations were calculated to examine zero order relationships between all variables (see Table 3.1). As expected, there were significant positive correlations among the central constructs of the EPPM. Sun-safe behavioral expectation was positively correlated with susceptibility (r = .34, p < .001), severity (r = .57, p < .001), self-efficacy (r = .72, p < .001), and response efficacy (r = .61, p < .001). A contrast variable was created comparing the UV skin damage visual (1) to all other visual conditions combined (0) (labeled, UV vs. All). UV vs. All was positively correlated with fear (r =.09, p < .001) and fear was also positively correlated with behavior expectation (r =.22, p < .001). 51 3.4.2 UV vs. All (H1 & H2) The first set of hypotheses postulated that a UV skin damage visual would generate greater (a) susceptibility, (b) severity, and (c) fear compared to all other visual categories combined (i.e., UV vs. All). Analysis of variance (ANOVA) was utilized to test this hypothesis. Compared to all other conditions combined, the UV skin damage visual did not generate greater susceptibility, F(1, 2199) = .06, p = .81, or severity, F(1, 2199) = .32, p = .57. However, UV vs. All generated greater fear, F(1, 2199) = 16.68, p < .001. A pairwise comparison showed that the UV skin damage condition produced significantly greater fear (M = 3.07, SE = .08) than all other visuals combined (M = 2.69, SE = .04, p < .001, Cohen’s d = .27), supporting H1c (means and standard errors are provided in Table 3.2). The second set of hypotheses postulated that a UV efficacy visual condition would increase (a) self-efficacy and (b) response efficacy, when compared to a control or text-only efficacy condition (see Table 3.2). ANOVA analysis demonstrated that the efficacy conditions did not differ for response efficacy, F(3, 2199) = .67, p = .57, but there was a marginally significant difference for self-efficacy, F(3, 2199) = 2.41, p = .07. Pairwise comparisons revealed that the UV efficacy visuals increased self-efficacy (M = 5.31, SE = .06) compared to the control efficacy condition (M = 5.06, SE = .06, p < .01, Cohen’s d = .11), text-only efficacy condition (M = 5.12, SE = .06, p = .02, Cohen’s d = .08), and UV efficacy visual + text condition (M = 5.12, SE = .06, p = .02, Cohen’s d = .08). There were no significant differences within types of efficacy condition for response efficacy. Thus, there was support for H2a, but not support for H2b. 52 3.4.3 Fear as a Mediator (H3) The third hypothesis stated that fear would mediate the relationship between exposure to the UV skin damage visual and behavioral expectations such that those in the UV condition would report greater fear, which would increase behavior expectations. Simple mediation analysis (PROCESS Model 4, see Hayes, 2018) was used to test this hypothesis. Consistent with H3, simple mediation analysis revealed that fear significantly mediated the relationship between exposure to UV skin damage visuals and behavioral expectation, effect = .07, Boot SE = .02, 95% Boot CI: .0373, .1086, effect size k2 (kappa-squared) = .02 (see Table 3.3). The UV skin damage visual condition generated greater fear (coefficient = .38, SE = .09, t = 4.06, p < .001), which then significantly triggered sun-safe behavior expectations (coefficient = .18, SE = .02, t = 10.72, p < .001). The simple mediation models are presented in Figure 3.1. 3.4.4 UV Efficacy as a Moderator (H4) Moderated mediation analysis (PROCESS Model 8, see Hayes, 2018) was used to test H4, which postulated that the UV efficacy visual would moderate the indirect path through fear such that the indirect effect of UV skin damage visuals on sun-safe behavior expectations would be larger for individuals receiving the visual efficacy. Moderated mediation analysis revealed that the UV efficacy visual condition significantly moderated the indirect path through fear (see Figure 3.2). However, contrary to H4, the indirect effect was significant in the control (explaining 4% of the variance in behavior expectations) and text-only (explaining 3% of the variance) conditions. In those conditions, the UV threat condition increased fear, which increased behavior 53 expectations. Readers might question whether this finding is an artifact of the non-UV condition containing UV imagery in the visual and visual + text efficacy conditions (i.e., the UV efficacy visual). If this explanation was valid, then fear would increase in the non-UV conditions when the UV efficacy visual was present. The data do not support this explanation as fear did not increase in the non-UV conditions when the UV efficacy visual was present (see Supplemental Material 3.6.9). Instead, fear decreased in the UV threat condition when the UV efficacy image was present. 3.4.5 Comparison of UV to Other Naturally Occurring Visual Categories (RQ1) RQ1 queried if there was another naturally occurring skin cancer risk visual type that yields equivalent or greater impact on (a) susceptibility (b) severity, (c) self-efficacy, (d) response efficacy, (e) fear, and (f) behavior expectations. A series of 5 (visual conditions) × 3 (replications) × 4 (efficacy conditions) between-participant ANOVAs were conducted with replications as a nested factor. Visual condition was significantly related to self-efficacy, F(4, 2190) = 3.06, p = .02, and fear, F(4, 2190) = 12.20, p < .001. For self-efficacy, sunburn visuals (M = 4.97, SE = .06) were significantly different than sun exposure visuals (M = 5.20, SE = .06, p < .01, Cohen’s d = .18), aging visuals (M = 5.23, SE = .06, p < .01, Cohen’s d = .21), and mole removal visuals (M = 5.23, SE = .06, p < .01, Cohen’s d = .21). None of the visuals were significantly different in producing self-efficacy compared to UV skin damage visual. For fear, the UV skin damage visual generated significantly greater fear (M = 54 3.07, SE = .08) when compared with the sunburn visual (M = 2.61, SE = .08, p < .001, Cohen’s d = .27), sun exposure visual (M = 2.39, SE = .08, p < .001, Cohen’s d = .40), and aging visual (M = 2.76, SE = .08, p < .01, Cohen’s d = .18), but not with the mole removal visual (M = 3.03, SE = .08, p = .72, Cohen’s d = .02). Thus, only the mole removal visuals produced equivalent fear as the UV sun damage visuals (see Table 3.5). Given this finding, readers might be interested to know how mole removal images compare to other images in general. Results of ANOVAs with the contrast Mole vs. All are presented in Table 3.2 and the simple mediation with fear and other EPPM variables are presented in Supplemental Material 3.6.11. When compared with all other visual conditions combined, mole removal visuals produce greater fear, and fear mediates the relationship between condition and behavior expectations. Thus, Mole vs. All yields the same pattern of results as UV vs. All. There was no significant interaction between visual and efficacy conditions on any of the dependent variables. Means, standard errors, and confidence intervals for interactions are presented in Supplemental Material 3.6.8. Some readers might be interested in an analysis that examines visual condition as a five-level categorical predictor. An output for simple mediation analyses with all five visual condition as a categorical predictor, fear as mediator, and behavior expectations as outcome is included in Supplemental Material 3.6.12. 3.5 Discussion The results demonstrate that stock UV skin damage visuals elicit fear, which in turn triggers positive sun-safe behavior expectations. Thus, it appears that stock UV skin 55 damage visuals are best categorized as fear appeals. Mole removal visuals demonstrated a similar pattern (increased fear and fear as a mediator). UV skin damage visuals triggered fear, but did not increase threat susceptibility, which raises questions about the relationship of fear and susceptibility. In past studies, personalized UV photos have increased perceived susceptibility (Emmons et al., 2011; see McWhirter & Hoffman-Goetz, 2013). Generally, and perhaps somewhat surprisingly, fear appeal research has not examined the relationship between fear and susceptibility (Tannenbaum et al., 2015). The research community would benefit from more fear appeal studies that examine the relationship between these two variables, and that test both as potential mediators. Researchers should also carefully consider how both constructs are measured. Indeed, studying and refining key constructs in fear appeal research is a crucial next step. The use of physiological measures—such as galvanic skin response (GSR), facial expression analysis and electroencephalogram (EEG)—may also help researchers to explicate measurement of, and the relationship between, fear and susceptibility. Fear mediated the relationship between stock UV imagery and sun-safe behavior expectations. This finding suggests that researchers should continue to explore the position of fear in the fear appeal research. For example, it is possible that fear appeals exert influence on outcomes by increasing fear directly. Researchers should engage the fear hypothesis by testing fear as a mediator. However, I should note that fear was the only affective construct measured in the study and the average fear produced was below the midpoint on a seven-point scale. This suggests that fear in public health campaigns might be relatively modest or, in a larger sense, that there might be other emotions in play. Researchers have reported the association of fear appeals with other emotions such 56 as hope (Nabi & Myrick, 2018) or mixture of sequential emotions such as sadness, fear, joy, and relief (Carrera, Muñoz, & Caballero, 2010). Thus, future fear appeal studies should include self-reported as well as physiological measures of multiple discrete emotions so as to explore the role of different discrete emotions in fear appeal message processing. Two types of visuals were found to generate more fear than others: UV skin damage and mole removal. One image within the mole removal category deserves additional commentary. Image 13 (see Supplemental Material 3.6.5) depicts a young girl with a particular gruesome wound. This image is unaltered—it has not been visually manipulated—and it has been used, and found to be effective at generating fear, by other researchers studying skin damage visuals (Mays & Zhao, 2016). Future research seeking a useful comparison visual to develop and refine alternative visual approaches would be well-advised to consider image 13, notably in situations where replications are challenging. Future studies should also examine mole removal and UV photos with a range of discrete emotions and other message variables such as defensive avoidance, memorability, and novelty to explicate the similarities and differences in impacts that these two categories of visuals create. In the current study, UV efficacy visual appeal generated higher self-efficacy compared to the combination of UV visual and text. One possible explanation is that the text message could have increased the cognitive demands of the message as it recommended several sun-safe behaviors, whereas the UV efficacy visual focused on only one behavior—sunscreen use (see Supplemental Materials 3.6.6 and 3.6.7). Followup studies should investigate the veracity of this explanation as well as the underlying 57 mechanisms at play. Moreover, the results also question the moderating role of efficacy messages. Here, I found that, contrary to the proposed hypothesis, the indirect effect through fear was larger in the control and text-only conditions. The exposure to efficacious visual messages generated higher efficacy, which then appeared to nullify fear; without that fear, participants had lower sun-safe behavior expectations. EPPM assumes that threat message will be effective when there is sufficient efficacy but our analysis suggests that efficacy and threat appeals may interact in different ways – sometimes it might be the case that efficacy messages diffuse threat messages. If true, this has significant implications for fear appeal research. The question becomes, do we want to eliminate fear, or is there a certain level of fear that we want to maintain? Should efficacy temper fear, as opposed to remove it completely? This might also be a dosage effect. The exposure of one UV photo generates fear because it is novel, but seeing the second UV photo (i.e., the UV efficacy visual) might have diffused that novelty aspect and diminished the fear. These questions need to be answered in future studies by evaluating the interaction of multiple forms, doses, and types of efficacy messages in subsiding fear and impacting behavioral outcomes (e.g., see Carcioppolo et al., 2013). Another interesting finding of the study is the effect of sunburn visuals on selfefficacy. Participants in the sunburn visuals condition reported significantly less selfefficacy compared to other visuals except the UV skin damage visual (Table 3.4). This suggests participants might have believed it is easier to prevent aging and mole removal scars compared to sunburn. One of the striking difference between sunburn visuals and aging and mole removal visuals is the temporality of the threat (Shipp & Aeon, 2018). UV and sunburn visuals depicts immediate damage to the skin, whereas aging and mole 58 removal visuals show the threat that can happen in the future. It might be the case that people feel more in control of future threats because they are temporally distant. Future research should examine the effect of threat temporality on self-efficacy and sun-safe behavior. The current study has number of limitations, which may be addressed in future studies. First, sun-safe behavior expectations were measured instead of behavior, thus the effects seen here may not translate to actual behavior. I should also acknowledge that because the entire study was completed in a continuous series, this might have made the participants’ responses vulnerable to demand effects. Future studies could engage this limitation via a Solomon four group design experiment (McCambridge, Kypri, & Elbourne, 2014; Solomon, 1949). Moreover, the mediators and outcome variables were all measured at the same point in time (i.e., the posttest), which limits the ability to perform meaningful mediation analysis and might introduce biases in the results (Kline, 2015; Maxwell & Cole, 2007; Tate, 2015). Researchers could design a longitudinal study that explores the impact of a stock UV photo in promoting sun-safe behaviors over time to mitigate these biases. Some UV intervention studies (e.g., Mahler et al., 2013) have used spectrophotometry that provides an objective measure of skin color, which tracks the actual practice of sun-safe behavior. This could be a superior outcome measure compared to self-report, and a valuable addition to longitudinal evaluations. A second limitation is the use of self-reported fear measured at a single point in time. Future studies could utilize physiological measures to assess fear and examine how fear manifests and progresses as participants view UV photos. Third, I have only studied the impacts of stock UV photos in the context of sun-safe behavior; future studies can 59 study the impacts of UV photos in promoting other skin cancer prevention behaviors such as skin self-examination. Fourth, the visuals used in this study varied in terms of demographics (i.e., age, sex, race) of the model and I did not study the potential effect of such factors. Future studies can systematically vary the demographics of the models— and perhaps the amount of skin damage (Mahler, 2018)—to understand the impact this might pose. Fifth, our analyses studied the danger control process portion of the EPPM model, but future UV intervention studies could test the full theoretical model of EPPM (including the fear control process). Sixth, some readers might question how participants perceived the UV visual-only efficacy stimuli. That condition did not include explanatory text about the visual or sun-safe behavior. To engage this limitation, I examined thought listing from the 555 participants in the visual-only condition. Of those, only 15 participants expressed any confusion or negative thoughts about the visual. Still, it is possible that participants in that condition may have perceived the visual in unexpected ways given the lack of explanatory text. The current study adds to the literature on UV photo interventions, the EPPM, and visual fear appeals because the findings demonstrate the potential of stock UV visuals in eliciting fear responses that lead to sun-safe behavioral practices. This has a practical implication in the use of stock UV visuals in the promotion materials, websites, and social media pages of organizations working to prevent skin cancer. Indeed, stock images like those examined here are commonplace in social media messages, and have been shown to increase attention and response (Vraga, Bode, & Troller-Renfree, 2016). Our findings also have an important implication for fear appeal models such as the EPPM, and for research on visual communication. To that end, the current study utilized a design 60 (multiple message categories with replications) that is rarely encountered in behavioral research (Jensen, 2008). Such designs require relatively large samples, but they also afford researchers numerous analytical opportunities. The current analysis examined the value of stock UV visuals compared to four alternative visual types. Other researchers may find value in carefully examining the other visuals and their impact. A larger goal of the study was to create a resource—including the visuals, data, and approach—to inform and support future research on visual communication, UV visuals, skin cancer, and fear appeals. Table 3.1 Bivariate Correlations 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12 1. Susceptibility 2. Severity ---.54*** 3. Self-efficacy .42*** ---.66*** 4. Response efficacy .46*** .72*** ---.79*** 5. .28*** .18*** .20*** ---.15*** ---- .34*** .57*** .72*** .61*** .22*** 7. Fear Behavior expectation UV vs. All .01 .01 -.01 .00 .09*** ---.00 8. Age .04* .12*** .14*** .16*** -.07*** .07** ---.01 9. Sex .01 .09*** .09*** .10*** -.08*** .08*** .03 ----.08*** 10. Education .13*** .09*** .13*** .13*** .09*** .10*** -.03† .02 ----.06** 11. Income .16*** .13*** .19*** .19*** .15*** .14*** -.01 .03 .00 ---.44*** 12. White .23*** .12*** .12*** .14*** -.01 .07*** -.01 .19*** .03 .06** ---.15*** 13. Hispanic -.04† -.02 .00 -.04† .04† .04† -.01 -.21*** -.03 -.02 -.02 ----.17*** 14. BRAT index .37*** .18*** .17*** .17*** .18*** .16*** .03 .04† .00 .13*** .16*** .36*** 6. 13 ----.03 Note. UV vs. All is the UV skin damage visual condition compared with all other visual conditions combined (UV skin damage =1, all other conditions combined=0). †p < .10; *p < .05; **p < .01; ***p < .001 61 Table 3.2 Means and Standard Errors for Mediators and Outcomes by Visual and Efficacy Condition UV Vs. All Mole Vs. All Efficacy conditions Susceptibility Severity Self-efficacy Response Efficacy Fear UV Other Conditions combined Mole Other Conditions combined No Efficacy 4.62(.08) 4.60(.04) 5.55(.07) 5.51(.03) 5.12(.06) 5.16(.03) 5.47(.07) 5.47(.03) 3.07(.08)a 2.69(.04)b Behavior Expectatio n 5.21(.07) 5.23(.03) 4.64(.08) 4.60(.04) 5.56(.07) 5.51(.03) 5.23(.06) 5.13(.03) 5.58(.07) 5.45(.03) 3.03(.08)a 2.71(.04)b 5.31(.07) 5.21(.03) 4.62(.07) 5.51(.06)a 5.06(.06)a 5.41(.06) 2.73(.07) 5.14(.06)a Text 4.60(.07) 5.51(.06)a 5.12(.06)a 5.50(.06) 2.80(.07) 5.17(.06)a b b Visual 4.64(.07) 5.61(.06)a 5.31(.06)b 5.55(.06) 2.76(.07) 5.35(.06)b Visual + Text 4.57(.07) 5.44(.06)b 5.12(.06)a 5.44(.06) 2.78(.07) 5.25(.06)ab Note. Means and standard errors (in parentheses). Means with different superscripts are significantly different, p < .05. 62 Table 3.3 Simple Mediation – Tests of Indirect Effects of EPPM Variables N = 2220 Models without mediator Models with mediator Bootstrap results for indirect effects (95% CI) Bootstrap results for indirect effect sizes (95% CI) Lower Upper k2 Lower Upper .01 −.0473 .0582 .01 .0000 .0065 .60*** .03 −.0621 .1153 .01 .0001 .0288 −.04 .78*** −.03 −.1392 .0790 .01 .0001 .0399 −.01 −.01 .64*** −.01 −.0995 .0947 .00 .0000 .0026 -.09 .38*** .18*** .0373 .1086 .02 .0105 .0305 B B R2 c R2 c’ Susceptibility .00 −.02 .12*** −.02 .02 .31*** Severity .00 −.02 .33*** −.04 .04 Self-Efficacy .00 −.02 .52*** .02 Response Efficacy .00 −.02 .37*** Fear .00 −.02 .05*** a b ab .07* 63 Note. Process Model 4 with 1000 bootstraps where each mediator was tested one at a time. Fear is the only significant mediator as the boot confidence interval does not overlap zero. Predictor is UV vs. All contrast, outcome is behavior expectation; B = unstandardized regression weights; c = total effect of predictor on outcome without the mediator in the model; c’ = direct effect of predictor on outcome while controlling for the mediator; a = the path between the predictor and the mediator; b = the path between the mediator and the outcome; ab = indirect effect of predictor on outcome thorough the mediator; R2 = amount of variance explained by the model; CI = confidence intervals; k2 = effect size. *p < .05; ***p < .001 Table 3.4 Means and Standard Errors of Mediators and Outcomes by Visual Condition Susceptibility Severity Self-Efficacy Response Efficacy Fear Behavior Expectation Sun 4.47(.08)a 5.50(.07) 5.20(.06)a 5.48(.07)ab 2.39(.08)a 5.27(.07) Exposure Sunburn 4.69(.08)b 5.47(.07) 4.97(.06)b 5.37(.07)a 2.61(.08)ab 5.12(.07) Visual ab a ab bc Aging 4.62(.08) 5.51(.07) 5.23(06) 5.47(.07) 2.76(.08) 5.23(.07) conditions Mole 4.64(.08)a 5.56(.07) 5.23(.06)a 5.58(.07)b 3.03(.08)d 5.31(.07) Removal UV 4.62(.08)ab 5.55(.07) 5.12(.06)ab 5.47(.07)ab 3.07(.08)d 5.22(.07) Note. Means and standard errors (in parentheses). Means with different superscripts are significantly different, p < .05. 64 65 Susceptibility UV vs. All -.02(.07) Behavioral Expectation -.02(.08) Severity UV vs. All -.04(.06) Behavioral Expectation -.02(.08) Self-Efficacy UV vs. All .02(.05) Behavior Expectation -.02(.08) Response Efficacy UV vs. All -.01(.06) Behavior Expectation -.02(.08) Fear UV vs. All -.09(.08) Behavior Expectation -.02(.08) Figure 3.1 Simple mediation model of the EPPM. Note. Only fear significantly mediated the relation between X (UV vs. All) and Y (Behavior Expectation): effect = .07, Boot SE = .02, 95% Boot CI: .0373, .1086, effect size (k2) = .02. ***p<.001 UV vs. All No Efficacy .74 (.19)*** Text Efficacy .45 (.19)* Visual Efficacy .12(.19) Visual+Text Efficacy .20(.19) Fear .18(.02)*** Behavior Expectation -.13(.15) Figure 3.2. Moderated mediation – indirect effect at four levels of efficacy condition. Note. Process Model 8 with 1000 bootstraps. The indirect path through fear was significant in the first two conditions (control and text), but it was no longer significant in the visual and visual + text condition. Control: Effect = .14, SE = .04, 95% CI = (.0692, .2099), k2 = .04a, SE = .01, 95% CI for k2 = (.0220, .0708) Text: Effect = .08, SE = .04, 95% CI = (.0182, .1536), k2 = .03a, SE = .01, 95% CI for k2 = (.0068, .0507) Visual: Effect = .02, SE = .03, 95% CI = (-.0461, .0877), k2 = .01bc, SE = .01, 95% CI for k2 = (.0001, .0167) Visual + Text: Effect = .04, SE = .04, 95% CI = (-.0322, .1064), k2 = .01bc, SE = .01, 95% CI for k2 = (.0007, .0282) Pairwise contrasts were used to compare conditional indirect effects. k2 that do not share superscripts are significantly different, p < .05. Notably, the conditional effect for control (k2 = .04) is significantly different than visual (k2 = .01) and visual+text (k2 = .01). *p < .05; ***p < .001 66 67 3.6 Supplemental Materials 3.6.1 UV Skin Damage Visuals Image 1. Image 2. Image 3. 68 3.6.2 Sun Exposure Visuals Image 4. Image 5. Image 6. 69 3.6.3 Sunburn Visuals Image 7. Image 8. Image 9. 70 3.6.4 Photo Ageing Visuals Image 10. Image 11. Image 12. 71 3.6.5 Mole Removal Visuals Image 13. Image 14. Image 15. 72 3.6.6 Text Efficacy Condition Stimuli There are a number of things that you can do to reduce your risk of skin cancer, including: • Wearing sunscreen • Staying out of the sun between 10 AM and 4 PM Wearing protective clothing (e.g., long sleeves, long pants, a broad brimmed hat, sunglasses) • Staying in the shade Next we're going to ask you a bit about the image you just saw. 73 3.6.7 Visual Efficacy Condition Stimuli 3.6.8 Estimated Marginal Means and 95% Confidence Intervals: Visual Conditions × Efficacy Conditions Dependent Variables Visual Condition Efficacy Condition Susceptibility No Efficacy Text Visual Visual + Text No Efficacy Sunburn Text Visual Visual + Text No Efficacy Aging Text Visual Visual + Text No Efficacy Mole Removal Text Visual Visual + Text No Efficacy UV Text Visual Visual + Text Self Efficacy Response Efficacy Fear Behavior Expectation Mean 95% CI Mean (SE ) 95% CI Mean (SE) (SE ) 4.24(.15) (3.95,4.54) 5.29(.13) (5.03,5.55) 4.92(.13) (4.67, 5.17) 5.26(.13) (5.00,5.52) 2.24(.16) 4.58(.15) (4.28,4.87) 5.54(.13) (5.28,5.79) 5.35(.13) (5.10, 5.59) 5.67(.13) (5.41,5.92) 2.26(.16) 4.72(.15) (4.42,5.01) 5.79(.13) (5.53,6.04) 5.48(.13) (5.24, 5.73) 5.68(.13) (5.43,5.94) 2.47(.16) 4.32(.15) (4.03,4.62) 5.40(.13) (5.15,5.66) 5.07(.13) (4.82, 5.31) 5.31(.13) (5.05,5.57) 2.58(.16) Mean 95% CI (SE ) (1.92, 2.56) 5.15(.14) (4.88,5.42) (1.94, 2.59) 5.32(.14) (5.05, 5.59) 4.67(.15) (4.37,4.97) 5.39(.13) (5.13,5.64) 4.80(.13) (4.55, 5.04) 4.70(.15) (4.40,4.99) 5.65(.13) (5.39,5.91) 5.00(.13) (4.75, 5.25) 4.63(.15) (4.33,4.93) 5.35(.13) (5.09,5.60) 5.08(.13) (4.83, 5.32) 4.75(.15) (4.46,5.05) 5.47(.13) (5.22,5.73) 5.01(.13) (4.76, 5.26) (2.12, 2.76) 5.00(.14) (4.73,5.27) (2.34, 2.99) 5.07(.14) (4.80, 5.34) Mean (SE ) Sun Exposure Severity 95% CI Mean (SE ) 95% CI 5.25(.13) (5.00,5.51) 2.44(.16) 5.53(.13) (5.27,5.79) 2.67(.16) 5.28(.13) (5.02,5.54) 2.73(.16) 95% CI (2.15, 2.80) 5.53(.14) (5.26, 5.80) (2.26, 2.90) 5.08(.14) (4.81, 5.35) (2.40, 3.05) 5.20(.14) (4.93, 5.47) (2.27, 2.91) 5.22(.14) (4.95, 5.49) 5.43(.13) (5.17,5.69) 2.59(.16) 4.44(.15) (4.14,4.73) 5.57(.13) (5.31,5.83) 5.21(.13) (4.96, 5.46) 5.40(.13) 4.74(.15) (4.45,5.04) 5.39(.13) (5.14,5.65) 5.06(.13) (4.81, 5.31) 5.35(.13) 4.72(.15) (4.43,5.02) 5.69(.13) (5.43,5.94) 5.48(.13) (5.23, 5.73) 5.62(.13) 4.57(.15) (4.27,4.86) 5.40(.13) (5.14,5.65) 5.17(.13) (4.92, 5.41) 5.50(.13) (5.14,5.66) 2.38(.16) (5.09,5.61) 2.87(.16) (5.36,5.88) 2.85(.16) (5.24,5.76) 2.93(.16) 4.93(.15) (4.63,5.22) 5.61(.13) (5.35,5.86) 5.28(.13) (5.03, 5.53) 4.48(.15) (4.18,4.78) 5.45(.13) (5.20,5.71) 5.15(.13) (4.90, 5.40) 4.56(.15) (4.26,4.85) 5.61(.13) (5.35,5.87) 5.19(.13) (4.94, 5.44) 4.59(.15) (4.29,4.88) 5.57(.13) (5.31,5.82) 5.29(.13) (5.04, 5.54) 5.60(.13) (5.34,5.86) 3.28(.16) 5.54(.13) (5.28,5.80) 3.05(.16) (2.96, 3.60) 5.22(.14) (4.95,5.49) (2.73, 3.37) 5.17(.14) (4.90, 5.44) 5.61(.13) (5.35,5.87) 2.91(.16) (2.59, 3.23) 5.38(.14) (5.11, 5.65) 5.55(.13) (5.29,5.81) 2.88(.16) 4.80(.15) (4.51,5.10) 5.69(.13) (5.43,5.94) 5.07(.13) (4.82, 5.32) 4.50(.15) (4.20,4.79) 5.52(.13) (5.27,5.78) 5.06(.13) (4.81, 5.31) 4.56(.15) (4.27,4.86) 5.61(.13) (5.36,5.87) 5.29(.13) (5.04, 5.54) 4.63(.15) (4.33,4.93) 5.61(.13) (5.12,5.64) 5.05(.13) (4.80, 5.30) 5.53(.13) (5.27,5.79) 3.33(.16) (2.55, 3.20) 5.45(.14) (5.18, 5.72) (3.00, 3.65) 5.15(.14) (4.88,5.41) 5.39(.13) (5.13,5.65) 3.16(.16) 5.57(.13) (5.31,5.82) 2.86(.16) (2.84, 3.48) 5.27(.14) (5.00, 5.54) (2.53, 3.18) 5.25(.14) (4.98, 5.52) 5.39(.13) (5.13,5.64) 2.94(.16) (2.62, 3.26) 5.20(.14) (4.93, 5.47) Note. Means and standard errors (in parentheses). (2.06, 2.71) 5.17(.14) (4.90,5.44) (2.55, 3.19) 5.04(.14) (4.77, 5.30) (2.53, 3.17) 5.41(.14) (5.14, 5.67) (2.60, 3.25) 5.31(.14) (5.04, 5.58) 74 3.6.9 Estimated Marginal Means and 95% Confidence Intervals for Fear: UV vs. All × Efficacy Conditions Visual Condition Non-UV Conditions UV Condition Efficacy Condition Control Text Visual Visual + Text Control Text Visual Visual + Text Mean(SE) 95% CI 2.59(.08) 2.71(.08) 2.74(.08) 2.74(.08) 3.33(.16)a 3.16(.16)ab 2.86(.16) b 2.94(.16)b (2.42, 2.75) (2.55, 2.87) (2.58, 2.90) (2.58, 2.90) (3.00, 3.65) (2.84, 3.49) (2.53, 3.18) (2.62, 3.26) Note. Means and standard errors (in parentheses). Means with different superscripts are significantly different, p < .10. In Non-UV conditions, the means are not significantly different. In UV conditions, fear in the control efficacy condition is significantly greater than in the visual efficacy condition, p = .04 and approaching significance in the Visual +Text efficacy condition, p = .098. 75 3.6.10 Simple Mediation – Tests of Indirect Effects of EPPM Variables (Mole vs. All) N = 2220 Models without mediator Models with mediator Bootstrap results for indirect effects (95% CI) Bootstrap results for indirect effect sizes (95% CI) ab Lower Upper k2 Lower Upper .31*** .01 −.0400 .0654 .00 .0000 .0116 .05 .60*** .03 −.0570 .1144 .01 .0003 .0381 .02 .10 .78*** .08 −.0273 .1876 .03 .0018 .0711 .38*** .02 .13† .64*** .08 −.0144 .1683 .03 .0025 .0567 .05*** .04 .32*** .18*** .06* .0233 .0963 .02 .0068 .0273 B B R2 c R2 c’ a b Susceptibility .00 .10 .12*** .08 .04 Severity .00 .10 .33*** .07 Self-Efficacy .00 .10 .52*** Response Efficacy .00 .10 Fear .00 .10 Note. Process Model 4 with 1000 bootstraps where each mediator was tested one at a time. Fear is the only significant mediator as the boot confidence interval does not overlap zero. The predictor is the Mole vs. All contrast, outcome is behavior expectation; B = unstandardized regression weights; c = total effect of predictor on outcome without the mediator in the model; c’ = direct effect of predictor on outcome while controlling for the mediator; a = the path between the predictor and the mediator; b = the path between the mediator and the outcome; ab = indirect effect of predictor on outcome thorough the mediator; R2 = amount of variance explained by the model; CI = confidence intervals; k2 = effect size. †p < .10; * p < .05; ***p < .001 76 77 3.6.11 Simple Mediation Models of the EPPM Susceptibility Mole vs. All .08(.07) .10(.08) Behavior Expectation Severity Mole vs. All .07(.06) .10(.08) Behavior Expectation Self-Efficacy Mole vs. All .02(.05) .10(.08) Behavior Expectation Response Efficacy Mole vs. All .02(.06) Behavior Expectation .10(.08) Fear Mole vs. All -.09(.08) Behavior Expectation .10(.08) Only fear significantly mediated the relation between X (Mole vs. All) and Y (Behavior Expectation): effect = .06, Boot SE = .02, 95% Boot CI: .0233, .0963, Effect size (k2) = .02 †p <.10; ***p < .001. 78 3.6.12 PROCESS Model 4 Simple Mediation Analysis Output With Visual Conditions as Predictor, Fear as Mediator, and Behavior Expectations as Outcome Run MATRIX procedure: **************** PROCESS Procedure for SPSS Version 3.00 ***************** Written by Andrew F. Hayes, Ph.D. www.afhayes.com Documentation available in Hayes (2018). www.guilford.com/p/hayes3 ********************************************************************** **** Model : 4 Y : Behavior Expectation X : Visual Conditions M : Fear Sample Size: 2220 Coding of categorical X variable for analysis: Visual Conditions Sun Exposure Sun Burn Ageing Mole Removal UV Skin dam X1: X2: X3: X4: X1 .000 1.000 .000 .000 .000 X2 .000 .000 1.000 .000 .000 X3 .000 .000 .000 1.000 .000 X4 .000 .000 .000 .000 1.000 Sun Burn Visuals contrast with Sun Exposure Visuals Ageing Visuals contrast with Sun Exposure Visuals Mole Removal Visuals contrast with Sun Exposure Visuals UV Skin Damage Visuals contrast with Sun Exposure Visuals ********************************************************************** **** OUTCOME VARIABLE: Fear Model Summary R p .1467 .0000 Model ULCI constant 2.5503 X1 .4449 X2 .5966 R-sq MSE F df1 df2 .0215 3.0094 12.1749 4.0000 2215.0000 coeff se t p LLCI 2.3889 .0823 29.0169 .0000 2.2274 .2166 .1164 1.8603 .0630 -.0117 .3682 .1164 3.1628 .0016 .1399 79 X3 .8683 X4 .9104 .6400 .1164 5.4971 .0000 .4117 .6821 .1164 5.8581 .0000 .4537 ********************************************************************** **** OUTCOME VARIABLE: Behavior Expectation Model Summary R p .2259 .0000 Model ULCI constant 4.9822 X1 .0019 X2 .0781 X3 .1042 X4 .0071 EPPMFear .2191 R-sq MSE F df1 df2 .0510 1.9920 23.8007 5.0000 2214.0000 coeff se t p LLCI 4.8279 .0787 61.3553 .0000 4.6736 -.1878 .0948 -1.9809 .0477 -.3737 -.1081 .0949 -1.1385 .2550 -.2943 -.0828 .0954 -.8682 .3854 -.2698 -.1801 .0955 -1.8871 .0593 -.3673 .1852 .0173 10.7118 .0000 .1513 ************************** TOTAL EFFECT MODEL **************************** OUTCOME VARIABLE: Behavior Expectation Model Summary R p .0427 .3992 Model ULCI constant 5.4050 X1 .0428 X2 .1506 R-sq MSE F df1 df2 .0018 2.0942 1.0131 4.0000 2215.0000 coeff se t p LLCI 5.2703 .0687 76.7381 .0000 5.1356 -.1477 .0971 -1.5205 .1285 -.3382 -.0399 .0971 -.4108 .6813 -.2304 - 80 X3 .2262 X4 .1366 .0357 .0971 .3677 .7131 -.1548 -.0538 .0971 -.5543 .5794 -.2443 ************** TOTAL, DIRECT, AND INDIRECT EFFECTS OF X ON Y ************** Relative total effects of X on Y: Effect se t c_ps X1 -.1477 .0971 -1.5205 -.1021 X2 -.0399 .0971 -.4108 -.0276 X3 .0357 .0971 .3677 .0247 X4 -.0538 .0971 -.5543 -.0372 p LLCI ULCI .1285 -.3382 .0428 .6813 -.2304 .1506 .7131 -.1548 .2262 .5794 -.2443 .1366 Omnibus test of total effect of X on Y: R2-chng F df1 df2 .0018 1.0131 4.0000 2215.0000 ---------Relative direct effects of X on Y Effect se t c'_ps X1 -.1878 .0948 -1.9809 -.1298 X2 -.1081 .0949 -1.1385 -.0747 X3 -.0828 .0954 -.8682 -.0572 X4 -.1801 .0955 -1.8871 -.1245 p .3992 p LLCI ULCI .0477 -.3737 -.0019 .2550 -.2943 .0781 .3854 -.2698 .1042 .0593 -.3673 .0071 Omnibus test of direct effect of X on Y: R2-chng F df1 df2 .0023 1.3194 4.0000 2214.0000 ---------- p .2604 Relative indirect effects of X on Y ImgCondi X1 X2 X3 X4 -> Effect .0401 .0682 .1185 .1263 EPPMFear BootSE .0212 .0219 .0238 .0233 -> BootLLCI .0004 .0273 .0742 .0841 POSTBEas BootULCI .0838 .1131 .1689 .1758 Partially standardized relative indirect effect(s) of X on Y: ImgCondi -> EPPMFear -> POSTBEas 81 X1 X2 X3 X4 Effect .0277 .0471 .0819 .0873 BootSE .0146 .0151 .0163 .0160 BootLLCI .0003 .0188 .0516 .0585 BootULCI .0574 .0780 .1161 .1207 *********************** ANALYSIS NOTES AND ERRORS ************************ Level of confidence for all confidence intervals in output: 95.0000 Number of bootstrap samples for percentile bootstrap confidence intervals: 5000 ----- END MATRIX ----- CHAPTER 4 VISUAL TAILORING AND SKIN CANCER PREVENTION: COMPARING PERSONALIZED, STOCK, AND NON-ULTRAVIOLET IMAGES Fear appeals utilize frightening information to influence human perception and behavior (Leventhal, 1971; Witte, 2013). Several theoretical frameworks have been proposed to study fear appeals, including drive theory (Hovland, Janis, & Kelley, 1953), protection motivation theory (Rogers, 1975), the parallel process model (Leventhal, 1970), and the extended parallel process model (EPPM; Witte, 1992). These theories have provided a strong platform for research supporting hundreds of studies across 7 decades (Witte, 2013). Despite sustained research attention, there are still significant unanswered questions at the center of fear appeal scholarship. Notably, Dillard et al. (2016) questioned the relationship between perceived fear and behavioral outcomes. A recent meta-analysis of fear appeal research echoed this question and noted a need for more empirical studies testing fear as a mediator: “empirical studies typically test the impact of fear appeal messages on outcomes, and subsequently infer that message effects were mediated by experienced fear even though fear itself is rarely measured” (Tannenbaum et al., 2015, p. 1180). Though often included as a manipulation check, Tannenbaum and 83 colleagues argued for more empirical tests of mediation. Put differently, researchers are still trying to explicate whether fear increases behavioral response (i.e., a positive linear relationship), decreases behavioral response (i.e., a negative linear relationship), or demonstrates a more nuanced pattern (e.g., a curvilinear relationship). Another central question driving fear appeal research is the identification of message features that consistently trigger arousal. For example, stimuli in fear appeal research often include a visual and sometimes even examine its impact (e.g., a graphic warning on tobacco packets, see White, Webster, & Wakefield, 2008). However, beyond studying the simple presence or absence of visuals lies an opportunity to consider more nuanced visual features (King, 2015a). Visuals come in two basic forms: preset images that are the same for all audiences (i.e., stock visuals) or those that are personalized for a particular message recipient (i.e., tailored visuals). Personalization or tailoring of a message based on an individual’s characteristics has been effective at changing behaviors in interventions and health campaigns (Kreuter, Strecher, & Glassman, 1999; Noar, Benac, & Harris, 2007). From a mechanism standpoint, some past studies have found that tailored messages generate greater fear compared to stock messages (Kessels, Ruiter, Brug, & Jansma, 2011). The current study engages both questions by examining fear response following exposure to tailored and stock visuals. Skin cancer prevention serves as the context for the study as it is outwardly visible and visuals are common in this domain. Moreover, sun-safe behavior interventions have examined the impact of ultraviolet (UV) photos, a form of personalized visual where participants view personal photos revealing skin damage as dark spots and patches. A meta-analysis found that exposure to UV photos 84 increases sun-safe behaviors (Williams et al., 2013). Accordingly, this study compares the relative impact of a personalized UV visual with a stock UV visual at invoking fear and promoting sun-safe behavior. 4.1 Studying Fear Appeals From a theory standpoint, the current study is situated in the EPPM (Witte, 2013). The central constructs of the EPPM are perceived threat, efficacy beliefs, and perceived fear (Witte, 1992; Witte et al., 1996). Perceived threat constitutes perceived susceptibility and perceived severity, where perceived susceptibility is one’s belief about the risk of being vulnerable to the harm/danger and perceived severity is the magnitude of seriousness of the harm. The EPPM further categorizes perceived efficacy beliefs as selfefficacy and response efficacy: self-efficacy is one’s ability to execute the recommended actions and response efficacy is the belief that the recommended actions are capable of preventing or averting the threats posed. The EPPM identifies three possible outcomes of message exposure: no response, message acceptance, or message rejection. A fear appeal is postulated to have impact only when the message communicates some amount of threat, notably susceptibility. In the absence of the threat component, individuals are mostly nonresponsive (Witte, 1992). However, message acceptance and message rejection are based on the interaction of perceived threat and efficacy beliefs. In other words, the motivation to engage in a recommended behavior is the outcome of the threat assessment and the perceived efficacy belief (Witte, 1994). Prior research indicates that high perception of severity and susceptibility together with high efficacy beliefs will activate a danger control process, 85 which leads to the acceptance of the recommended behavior (Witte et al., 1996). However, when the efficacy belief is not sufficient, fear is produced, which triggers a fear control process that leads to the rejection of message recommendations (Witte & Allen, 2000). The EPPM synthesizes earlier fear appeal theories and is one of the more common frameworks in current research (Maloney et al., 2011). Thus, the EPPM provides a well-known theoretical framework to evaluate the influence of UV photographs in promoting sun-safe behavior expectations 4.1.1 Skin Cancer Prevention, Ultraviolet (UV) Photo Interventions, and the EPPM Skin cancer is the most common type of cancer in the United States as evidenced by its escalating incidence rate (National Cancer Institute, 2016). The three most common types of skin cancer—basal cell carcinoma, squamous cell carcinoma, and melanoma— are primarily caused by exposure to sunlight containing UV rays (Gandini, Sera, Cattaruzza, Pasquini, Picconi, et al., 2005). Despite strong awareness of this risk factor, the most dangerous type of skin cancer – melanoma – continues to increase in incidence (Siegel et al., 2017). UV photographs or photos are taken with special cameras that filter out visible light (Fulton, 1997). As such, UV photos allow humans to see objects and their environment exclusively in UV light. UV photos or images are the only way people can truly see the full effects of UV rays as the human eye is not capable of viewing this spectrum (Fulton, 1997). Researchers have typically studies personalized UV images 86 (Williams et al., 2013) or stock UV images, but there is a pressing need for more studies comparing the relative impact of the two as presented in Chapter 3. UV photo interventions have demonstrated meaningful impact on skin cancer prevention behaviors. UV interventions have been found to avert intentions to tan, tanning behavior, and decrease actual skin tan (Gibbons et al., 2005; Mahler et al., 2007). Effective interventions have been carried out for children aged 11 to 13 years (Demierre et al., 2009) as well as older teenagers (Taylor et al., 2016). The photos are more than a novelty; the damage seen in UV photographs is strongly correlated with phenotypic melanoma risk (Gamble et al., 2012). From a mechanism standpoint, it appears that UV photos are effective at shaping behavior because the intervention depicts a threat. Consistent with this idea, UV photo interventions have been shown to impact behavior by increasing perceived susceptibility to skin cancer and ageing (Gibbons et al., 2005; Walsh & Stock, 2012; Walsh, Stock, Peterson, & Gerrard, 2014), negative emotional reactions (Mahler, 2014, 2018), and/or— as reported in Chapter 3—fear. 4.2 Visual Tailoring Tailoring is defined as “any combination of strategies and information intended to reach one specific person, based on characteristics that are unique to that person, related to the outcome of interest, and derived from an individual assessment” (Kreuter, Farrell, Olevitch, & Brennan, 2000, p. 277). Tailored communication is prepared uniquely for an individual based on a personal assessment. Tailoring health information has been found to increase information recall (Nguyen et al., 2017), satisfaction related to the health 87 information provided (Nguyen, Smets, Bol, Loos, & van Weert, 2018), screening intentions (Jensen et al., 2012), preventive behavior (Scholes et al., 2003), and has increased behavior for numerous interventions (Noar et al., 2007). Tailored messages often include a visual (Kreuter et al., 2000). Commonly, visuals are tailored based on demographic variables such as gender (e.g., Kreuter, Oswald, et al., 2000) and race/ethnicity (e.g., Scholes et al., 2003). Tailored print materials with visual elements are found to be more effective in influencing behaviors (Noar et al., 2007). Jensen et al. (2012) found that tailored illustrated pamphlets increased screening intentions for breast cancer. A meta-analysis of tailored health messages called for researchers to explicitly state how visuals are tailored to the specific individuals and formally examine the impact of visual tailoring (Noar et al., 2007). Tailoring holds great promise for communication interventions and practice, but only when it yields sufficient impact to warrant the additional labor required to personalize a message (Jensen et al., 2014). Personalized UV photos require specialized equipment and an opportunity for the photographer to take a photo of the participant. This additional labor needs to be weighed against the convenience of using stock UV images that are freely available. More to point, stock UV photos have been found to be effective at increasing sun-safe behaviors. The study presented in Chapter 3 compared stock UV photos to other naturally occurring visuals used in skin cancer promotion materials (i.e., visuals depicting sun exposure, sun burn, ageing, and mole removal) and observed that UV stock photos produced greater perceived fear when compared to other visual conditions. Perceived fear also mediated the relationship between exposure to the stock UV photos and sun-safe behavior expectations such that stock UV photos generated 88 greater fear (compared to other visuals) and increased behavior expectations. Chapter 3 highlighted that the majority of studies examining the impact of UV photos have used personalized images, but there is a need for additional research examining the relative impact of personalized and stock UV photos. In the present study, there are three visual conditions: a personalized UV (PUV) photo, a stock UV (SUV) photo, and non-UV (NUV) photo. The NUV photo depicts a person with mole removal scars. It is an original photo created by another research team (Mays & Zhao, 2016) and used in the aforementioned comparison of multiple visuals in chapter three. The mole removal visual has been found to produce fear (Mays & Zhao, 2016), including equivalent fear as the stock UV condition, as seen in Chapter 3. Thus, the mole removal image is an ideal comparison condition for this research program. H1: A personalized UV (PUV) skin damage photo will generate greater (a) susceptibility, (b) severity, (c) perceived fear, and (d) behavior expectations compared to (i) a stock UV skin damage (SUV) photo or (ii) a No UV skin damage (NUV) photo. The study presented in Chapter 3 targeted efficacy beliefs (i.e., self-efficacy and response efficacy) utilizing UV efficacy visuals. In the current study, the same approach is used but there are only two efficacy conditions: efficacy and no efficacy. Participants in the efficacy condition received one of the stimuli used in the aforementioned chapter. The stimuli depicts a UV visual of an individual half of whose face is covered by sunscreen (which appears dark in UV light) paired with some behavioral recommendations to prevent skin cancer. I hypothesize that the efficacy visual will generate greater efficacy beliefs and behavior expectations. 89 H2: Compared to the no efficacy message conditions, efficacy message visuals will generate greater (a) self-efficacy, (b) response efficacy, and (c) behavior expectations. 4.3 Emotion and Physiological Methods in Fear Appeal Research Perceived fear is the only emotion in the EPPM. Fear is postulated to invoke message rejection by activating a fear control process (Witte, 1992, 1994). However, scholars have questioned this postulate, with some arguing that fear can trigger both danger control and fear control (So, 2013) and others noting that fear may be more likely to drive behavior change, akin to the drive model (Dillard, Li, Meczkowski, et al., 2016). Concerning the latter, positive relationships between fear and persuasive outcomes have been evidenced (De Meulenaer, De Pelsmacker, & Dens, 2015), including in the context of stock UV photo interventions, as reported in Chapter 3 of this dissertation. Given that, current study hypothesizes the mediating role of fear. H3: Fear will mediate the relationship between exposure to UV skin damage visuals and sun-safety behavioral expectations, such that participants in the PUV skin damage visual condition will report greater fear, which will be positively related to behavioral expectations compared to participants in (a) SUV and (b) NUV conditions. However, questions about the role of fear in the EPPM are foregrounded by larger concerns about how to measure affect (Weidman, Steckler, & Tracy, 2017). That is, meaningful and reliable postulates about fear require meaningful and reliable measures of 90 fear. One of the ways to engage questions about the construct of fear is by utilizing physiological measures. Researchers have suggested measuring physiological response to fear appeal messages as a way to advance theory (Ordoñana, González-Javier, EspínLópez, & Gómez-Amor, 2009; Popova, 2012). In the context of UV photo interventions, Mahler (2014) made a similar argument when she suggested that future research should examine physiological responses to UV stimuli, and compare those responses with other cognitive measures. Researchers have reported that messages that generate greater selfreported fear produce higher physiological arousal (Lee & Lang, 2009) Thus, in the current study, I hypothesize that: H4: PUV skin damage photos will generate greater physiological arousal compared to (a) an SUV photo and (b) an NUV photo. H5: Physiological arousal will mediate the relationship between exposure to UV skin damage visuals and sun-safety behavioral expectations, such that participants in the PUV visual condition will report greater physiological arousal, which will be positively related to behavioral expectations regarding (a) sunscreen use, (b) protective items, and (c) seeking shade compared to participants in (i) SUV and (ii) NUV conditions. Researchers have argued that a range of emotions are related to behavior and behavior change (Moser, Mccaul, Peters, Nelson, & Marcus, 2007; Weinstein et al., 2007). Consistent with this idea, past research has suggested that several other emotional responses – surprise, puzzlement, anger, sadness, irritation, disgust, tension, anxiety – can be evoked by fear appeals in addition to fear (Dillard, Plotnick, Godbold, Freimuth, & Edgar, 1996; Passyn & Sujan, 2006). Emotions such as hope (Nabi & Myrick, 2018) or 91 mixture of series of emotions such as sadness, fear, joy, and relief (Carrera, Muñoz, & Caballero, 2010) have been associated with fear appeals. Thus, the current study measures positive and negative valence using facial expression analysis to explore if these affective domains vary by visual conditions and if any of these alternative emotional responses mediate the relationship between exposure to visual condition and behavioral expectations. RQ1: Do the positive and negative valence detected in facial expressions systematically vary by visual condition? RQ2: Do the positive and negative valence detected in facial expressions mediate the relation of exposure to visual condition and behavior expectations? 4.4 Methods 4.4.1 Study Design This is a 3 (visual conditions: PUV, SUV, NUV) × 2 (efficacy conditions: efficacy, no efficacy) between-participants message experiment. Participants in the PUV visual condition viewed their own UV photos, participants in the SUV visual condition saw a standard UV image, and participants in the NUV condition saw a non-UV photo of a girl with mole removal surgery. For the efficacy condition, half of the sample received an efficacy inducing message and the other half did not. 4.4.2 Participants and Procedure One hundred and eight undergraduate students (age range: 18-50, Mage = 23.10, SD = 5.77) enrolled in a Department of Communication course at a western university 92 were recruited for the study. Most participants were female (n = 72) and identified as Caucasian (n = 86). The remaining participants identified as follows: African American (n = 3), American Indian or Alaska Native (n = 2), Asian Indian (n = 2), Japanese (n = 1), Korean (n = 7), Filipino (n = 2), Vietnamese (n = 3), Samoan (n = 1), other Pacific Islander (n = 1), and some other race (n = 11). There were some participants who identified as Latino (n = 14). Based on the BRAT index, 59 participants were categorized as having a low risk, 31 participants as having a moderate risk, and 18 participants having a high skin cancer risk. They were offered extra-credit as a compensation for participation. The students filled out a consent form, completed an online pretest, and scheduled a 30-minute lab session. There was a gap of at least 15 days between the pretest and the lab session. Upon arrival to the lab, students were randomly assigned to one of six groups experimental conditions. Participants in the PUV condition had a picture of their face taken with a professional VISIA camera. The photo was then transferred to computer that tracks physiological responses. It is important to note that participants in the PUV condition could not see their photo while it was being taken. They only viewed the image later in the study to make it consistent with the other conditions. After taking a seat in front of the computer installed with iMotions software, a galvanic skin response (GSR) shimmer device was attached to their left hand. Following a text prompt (see Supplemental Material 4.7.1), participants were exposed to the respective intervention stimulus (i.e., the personalized UV visual, stock UV visual, or mole excision visual). They were allowed as much time as they would like to review the visual. Following the display of the visual condition, participants were exposed to the efficacy stimuli (or not). 93 Finally, the participants completed the posttest survey, which had questions about perceived fear, susceptibility, severity, self-efficacy, response efficacy, and behavior expectations. 4.4.3 Stimuli Participants in the PUV group were asked to not wear any make up including sunscreen and moisturizers. Before taking the picture, participants were offered to wipe their face with a soap-based wipe; some participants opted in and others opted out. Frontal view of the full-face was captured using a VISIA version 7 Facial Complexion Analysis System. The photos were stored on a password-protected folder in a lab computer. The SUV visual was produced using the same camera system. The UV photo depicted one of the study authors who did not assist with data collection and does not teach at the university where data were collected. The NUV visual is a mole excision image of a young girl that has been used in prior fear appeal research (Mays & Zhao, 2016) and also in Chapter 3. 4.4.4 Measures 4.4.4.1 Demographics Demographics including age, sex, and ethnicity were collected in the pretest. Skin cancer risk was assessed use the brief risk assessment tool (BRAT; Glanz et al., 2003). 94 4.4.4.2 Behavior Expectations Armitage and colleagues (2015) observed that expectations, as opposed to intentions, are a better predictor of behavior. Behavior expectations measure the likelihood of performing a behavior rather than simply the intention. Sun-safety behavior expectations were assessed using seven items measured on a seven-point scale ranging from 1 (extremely unlikely) to 7 (extremely likely) (Pretest: α = .78, M = 5.01, SD =1.14; Posttest: α = .84, M = 5.30, SD = 1.22). The measure was created by rewording fiveitems from a previously validated behavioral measure (Aspinwall et al., 2014) then adding two items. In the context of sun-safe behaviors, there are a few measurement discrepancies in the literature. Some researchers have used a combined sun-safe behavior measure (e.g., Manne et al., 2010) while others have treated individual sun-safe behaviors (e.g., application of sunscreen, use of protective item, staying in shade) separately (Aspinwall, Leaf, Kohlmann, Dola, & Leachman, 2009; Aspinwall et al., 2014; Miller et al., 2015). The current study used seven items to create three composite scales that accounts for three separate sun-safe behavior expectations, the logic being that each behavior is unique and might not always occur in combination. The individual items “using sunscreen” and “reapplied sunscreen after swimming or perspiration” were used to create a measure for behavior expectations related to sunscreen (Posttest: r = .82, M = 5.96, SD = 1.31). The items “wearing protective clothing (long pants and sleeves),” “wearing broad brimmed hat,” and “wearing sunglasses” created a measure for behavior expectations for use of protective items (Posttest: α = .65, M = 5.30, SD = 1.35). Finally, these two items “avoiding peak UVR exposure from 10 AM to 4 PM” and “stayed in the 95 shade” were used for the behavioral expectation measure related to seeking shade (Posttest: r = .74, M = 4.71, SD = 1.79). 4.4.4.3 Self Efficacy Self-efficacy was assessed using nine items from Witte (2000) measured on a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest: α = .70, M = 4.76, SD =.84; Posttest: α = .70, M = 4.75, SD = .91). The individual items were “I am able to use sunscreen with at least SPF-15 or higher to prevent skin cancer,” “Using sunscreen with at least SPF-15 or higher to prevent skin cancer is easy for me,” “Using sunscreen with at least SPF-15 or higher to prevent skin cancer is inconvenient for me,” “Applying sunscreen to all areas of my body exposed to the sun to prevent skin cancer is difficult for me,” “Reapplying sunscreen every two hours to prevent skin cancer is convenient for me,” “Reapplying sunscreen after swimming or perspiring to prevent skin cancer is easy for me,” “Wearing clothing that covers my body to prevent skin cancer is inconvenient for me,” “Wearing a hat that provides shade for my face to prevent skin cancer is easy for me,” and “I am able to minimize my exposure to the sun at midday to prevent skin cancer.” 4.4.4.4 Response Efficacy Response efficacy was assessed using eight items from Witte (2000) measured on a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest: α = .96, M = 5.76, SD = .98; Posttest: α = .94, M = 5.91, SD = .84). Response efficacy items include: “My using sunscreen is effective in preventing skin cancer,” “Applying 96 sunscreen with at least a SPF-15 or higher is effective in preventing skin cancer,” “Applying sunscreen to all areas of my body exposed to the sun is effective in preventing skin cancer,” and “Reapplying sunscreen every two hours is effective in preventing skin cancer.” 4.4.4.5 Susceptibility Susceptibility to developing skin cancer was measured using these three-items developed by Witte and colleagues (1996): “I am likely to get skin cancer sometime during my life”, “I am at risk of getting skin cancer sometime during my life,” and “It is possible that I will get skin cancer sometime during my life.” Items were assessed on a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest: α = .89, M = 4.58, SD =1.38; Posttest: α = .82, M = 4.86, SD = .93). 4.4.4.6 Severity Items developed by Witte et al. (1996) were used to measure severity. These items—“I believe that skin cancer is a severe health problem,” “I believe that skin cancer is a serious threat to my health,” and “I believe that skin cancer is a significant disease”— were measured on a seven-point scale ranging from 1 (strongly disagree) to 7 (strongly agree) (Pretest α = .82, M = 5.44, SD =1.14; Posttest α = .81, M = 5.81, SD = .92). 4.4.4.7 Fear Witte’s (2000) six-item scale was used to assess perceived fear produced by the study stimuli. The six items ask participants to rate how much the message made them 97 feel – “frightened,” “tense,” “nervous,” “anxious,” “uncomfortable,” and “nauseous” on a seven-point scale ranging from 1 (not at all) to 7 (very much) (α = .93, M = 3.65, SD = 1.60). 4.4.4.8 Physiological Arousal Physiological arousal was measured by using a wireless shimmer that measures galvanic skin response (GSR), or skin conductance. GSR evaluates the electro dermal activities by monitoring the sweat secretion (in the hands or feet) and provides data about the level of arousal produced when encountered with any stimulus. GSR reflects the activities of the sympathetic nervous system that are related with motivation (Bradley & Lang, 1988). The electrodes wrapped in Velcro straps were placed on the base of both the index and middle fingers of the nondominant hand. Participants were requested to keep the movement of the hands at a minimum, as physical mobility impacts data quality. The raw data of GSR was set to automatic detection of peaks by defining the threshold peak at .01uS. After this, the data for peaks per minute were exported to Excel. A pivot table was used in Excel to compute the average peaks per minute of each participant. Then the data were merged into an SPSS data file with other survey data for further statistical analyses. 4.4.4.9 Emotional Valence Emotional valence were measured using a facial expression analysis system provided by iMotions, which is a nonobtrusive, automated, and computer-based technique of evaluating movement of facial muscles and determining expressions as a 98 proxy to study the valence of emotions and feelings associated with those facial expressions. The algorithms, known as AFFDEX, classifies emotions based on facial expressions. AFFDEX provides data for positive and negative emotional valence as well as detects the seven basic emotions (anger, sadness, disgust, fear, joy, surprise, and contempt; iMotions, 2016). After completion of the study, the data were postprocessed to maintain optimum quality. During postprocessing, new variables were created from the raw data. Data for percentage of time participant’s experienced positive and negative emotional valence when observing the PUV, SUV, or NUV stimuli were exported to an Excel spreadsheet. A pivot table was used to compile the average percentage of time each participant exhibited negative and positive valence and these data were merged with other self-reported data in an SPSS file. 4.4.5 Randomization Check A two-way MANOVA (factors: visual, efficacy) was carried out to check if randomization held for demographics (i.e., age, education, income) and the BRAT index. The analysis showed nonsignificant multivariate tests for visual condition, Pillai’s Trace = .04, F(8, 200) = .54, p =.83, efficacy condition, Pillai’s Trace = .04, F(4, 99) = .90, p = .47, and the visual × efficacy interaction, Pillai’s Trace = .12, F(8, 200) = 1.54, p = .44. Next, I ran a two-way MANOVA to examine the randomization for pretest scores of EPPM constructs (susceptibility, severity, response efficacy, and self-efficacy) and three measures of sun-safe behavior expectation. The multivariate test was not significant for visual condition, Pillai’s Trace = .15, F(14, 194) = 1.14, p =.32, efficacy condition, Pillai’s Trace = .06, F(7, 96) = .83, p = .57, or the visual condition × efficacy interaction, 99 Pillai’s Trace = .14, F(14, 194) = 1.02, p = .44. Results from these two MANOVAs demonstrate that randomization was successful for the demographic variables, BRAT, pretest EPPM variables, and pretest behavior expectation measures. 4.4.6 Power Analysis Due to sample size limitations, it is important to report a power analysis to help readers contextualize the findings. G*Power was utilized to calculate posthoc power given an alpha of .05, sample size of 108, and 6 conditions (Erdfelder, Faul, & Buchner, 1996). The design had limited power to detect a small effect (f = .10, power = .14), modest power to detect a medium effect (f = .25, power = .62), and excellent power to detect a large effect (f = .40, power = .97). 4.5 Results 4.5.1 Bivariate Correlations Bivariate correlations were computed to analyze the zero order relationships between all variables (see Table 4.1). Consistent with past research, there were significant positive correlations between susceptibility and severity (r = .32, p < .001) and self-efficacy and response efficacy (r = .41, p < .001). Fear was significantly correlated with severity (r = .30, p < .001), self-efficacy (r = .18, p = .07), response efficacy (r = .25, p = .01, novelty (r = .49, p < .001), and behavioral expectations related to sunscreen use (r = .19, p = .047), and seeking shade (r = .17, p = .079). Novelty was positively correlated with behavioral expectations related to sunscreen use (r = .36, p < 100 .001), use of protective items (r = .27, p = .004), and seeking shade (r = .25, p = .008). Novelty was negatively correlated with physiological arousal (r = -.28, p = .003). Three contrast variables were created: SUVvsPUV (SUV coded as 0 and PUV as 1), NUVvsPUV (NUV coded as 0 and PUV as 1), and NUVvsSUV (NUV coded as 0 and SUV as 1). The contrast NUVvsPUV was significantly correlated with fear (r = .29, p = .01), positive valence (r = .30, p = .01), and novelty (r = .24, p = .047). The contrast SUVvsPUV was correlated with behavior expectations related to protective items (r = .29, p = .01) and the contrast NUVvsSUV was correlated with novelty (r = .33, p = .005). 4.5.2 Comparing PUV, SUV, and NUV (H1, H2, H4) H1 and H4 postulated that a PUV skin damage visual would generate greater susceptibility, severity, perceived fear, behavior expectations, and physiological arousal compared to both SUV and NUV skin damage visuals. Two-way ANOVAs demonstrated that susceptibility F(2, 102) = 2.09, p =.13, severity F(2, 102) = 1.85, p =.16, behavior expectations for sunscreen F(2, 102) = 1.03, p =.36, behavior expectations for seeking shade F(2, 102) = 1.17, p =.31, and physiological arousal F(2, 102) = .22, p =.81, were not significantly different across the three visual conditions. There was a significant difference for perceived fear, F(2, 102) = 3.49, p =.03, and a marginally significant difference for behavior expectations for protective items F(2, 102) = 2.90, p = .059. A pairwise comparison demonstrated that the PUV skin damage condition produced significantly greater fear (M = 4.17, SE = .27) than NUV (M = 3.19, SE = .26, p = .01), but not significantly greater than SUV (M = 3.61, SE = .26, p = .13). Thus, H1c-ii was supported (see Tables 4.2 and 4.3 for means and standard errors). PUV produced greater 101 behavior expectations (M = 5.72, SE = .23) for protective items compared to SUV image (M = 4.98, SE = .22, p = .02), but not compared to NUV (M = 5.21, SE = .22, p = .11). H2 stated that an efficacy visual would increase self-efficacy, response efficacy, and behavior expectations when compared to a no efficacy message condition. ANOVA demonstrated that the efficacy conditions did not differ for self-efficacy, F(1, 102) = 1.46, p =.23, response efficacy, F(1, 102) = 2.39, p = .13, and behavior expectation for sunscreen F(1, 102) = 1.44, p = .36, behavior expectation for protective items F(1, 102) = .88, p = .35, and behavior expectation for protective items F(1, 102) = .09, p = .77. Thus, there was no support for the H3 (see Table 4.2 for means and standard errors). 4.5.3 Fear and Physiological Arousal as Mediators (H3, H5) H3 and H5 postulated that fear and physiological arousal would mediate the relationship between exposure to the UV skin damage visual and behavioral expectation such that those in the PUV condition would report greater fear, which would increase behavioral expectation compared to the SUV and NUV conditions. Mediation analysis was conducted using PROCESS (Model 4, see Hayes, 2018). In light of the bivariate correlations between mediator variables, separate models were conducted for fear and physiological arousal (Preacher & Hayes, 2008). Two contrasts were utilized as independent variables (i.e., PUV vs. SUV and PUV vs. NUV) to test H3 and H5. Compared to the NUV condition, the PUV skin damage condition generated greater fear (coefficient = .98, SE = .39, t = 2.52, p = .01), which then significantly triggered sunscreen related behavior expectation (coefficient = .18, SE = .09, t = 2.08, p =.04). The simple mediation models for each of the analyses are presented in Figures 4.1 and 4.2. 102 4.5.4 Positive and Negative Valence (RQ1 and RQ2) RQ1 queried if the affective constructs (i.e., positive and negative valence) differed systematically by condition (i.e., PUV, SUV, and NUV). ANOVA showed a significant difference in the production of positive valence across the three skin damage conditions as recorded by facial expression analysis, F(2, 100) = 4.61, p =.01. A pairwise comparison demonstrated that the PUV skin damage condition produced significantly greater positive valence (M = 7.03, SE = 1.67) than NUV (M = .11, SE = 1.63, p = .04) and SUV (M = 2.17, SE = 1.63, p = .004) (means and standard errors in Table 4.3). However, the ANOVA test for negative valence showed that the three conditions did not significantly differ in the production of negative valence F(2, 100) = .41, p =.66. As a follow-up analysis of the positive valence findings, we further evaluated if the conditions varied in producing positive discrete emotions. ANOVA demonstrated that there was a significant difference in the production of joy across the three skin damage conditions, F(2, 100) = 4.07, p =.02. A pairwise comparison demonstrated that the PUV skin damage condition produced significantly greater joy (M = 7.33, SE = 1.79) than NUV (M = .37, SE = 1.74, p = .006) and SUV (M = 2.42, SE = 1.74, p = .05). Next, RQ2 queried if positive valence and negative valence played a mediating role in the relation between exposure to one of the skin damage visuals (PUV, SUV, and NUV) and behavioral expectation. Mediation analysis showed that neither positive valence nor negative valence significantly mediated the relationship between exposure to PUV skin damage visuals and behavioral expectation, as compared to SUV and NUV conditions, for all three types of behavior expectation variables (see Figure 4.3). 103 4.5.5 The Role of Novelty As UV photos are a novel visual, I conducted post-hoc analysis to evaluate if novelty significantly varied across experimental conditions. ANOVA demonstrated a significant difference in novelty across the three skin damage conditions, F(2, 102) = 4.51, p =.01, but no significant difference across the efficacy conditions, F(1, 102) = .23, p =.63. A pairwise comparison demonstrated that PUV evoked greater perceptions of novelty (M = 5.07, SE = .21) than NUV (M = 4.41, SE = .21, p = .03) but not SUV (M = 5.26, SE = .21, p = .52). The means and standard errors for novelty in each conditions are presented in Table 4.2. Next, the mediating role of novelty was tested using PROCESS Model 4 (Hayes, 2013). Mediation analysis demonstrated that novelty significantly mediated the relationship between exposure to PUV vs. NUV and SUV vs. NUV skin damage visuals and all three types of behavioral expectations (i.e., sunscreen, protective items, and seeking shade), such that exposure to a PUV or SUV compared to NUV skin damage visuals increased perceived novelty, which produced positive behavior expectations. The simple mediation models for each of the analyses with novelty as a mediator is presented in Figure 4.4. 4.6 Discussion The results of the current study demonstrated that personalized UV skin damage visuals elicit greater fear, positive valence, and novelty as compared to a non-UV mole excision visual. Taken together, the findings suggest that personalized or tailored UV images invoke numerous affective and perceptional responses, a complex cocktail of 104 emotions and thoughts that hints at the need for equivalent nuance in theory and measurement. In the next few pages, I consider these findings carefully to guide research moving forward. Contrary to the finding of the study reported in Chapter 3, fear did not mediate the relationship between exposure to the UV photos and behavior expectations. This result might be the manifestation of limited statistical power because of small sample size in each experimental conditions. Future studies should test this relationship in a larger sample size. It would also be interesting to investigate the role of fear in a series of UV photo experiments, where participants’ belief regarding the damage of other people their age is altered (see Mahler, 2018). Researchers have reported that fear appeals can trigger other emotions such as hope (Nabi & Myrick, 2018) or mixture of sequential emotions such as sadness, fear, joy, and relief (Carrera, Muñoz, & Caballero, 2010). Accordingly, I utilized physiological measures and found that PUV skin damage visuals produced significantly greater positive valence as manifested in facial expressions while viewing the stimuli. Thus, growing evidence suggests that a mixture of emotions might underlie response to fear appeals. However, significant questions remain. For example, was the positive valence best described as happiness related to viewing a UV photo? Or surprise about the extent of sun damage? Or hope that the sun damage could be prevented in future? As a next step, researchers could pursue a qualitative study with participants to explore the underlying thought processes and affective mechanisms. Another notable finding of this study is that fear was significantly different across the visual conditions but the physiological arousal, which should pick up on fear, did not 105 differ across groups. There are several possible explanations for this response pattern. First, it could be that the fear generated by these UV images is very subtle—indeed, the mean scores for fear are relatively low—and that may be insufficient intensity for a skin conductance measure to pick up (Aue, Hoeppli, & Piguet, 2012). The participants might have been concerned, but not alarmed. Future research should explore the minimum required intensity of self-reported fear that is captured by a GSR shimmer. Another possibility is that being in a laboratory environment might have made the participants more aware of the situation and hence physiological indicators were suppressed. Post hoc analyses demonstrated that message impact theory variables (Jensen et al., 2017), notably novelty, are important in understanding the impact of UV photos. This study found that novelty is one of the mechanisms through which personalized and stock UV photos impact positive behavioral outcomes. This finding has possible implications for message design as it suggests novel messages are influential in persuasion efforts. Future researchers can investigate how other variables of the message impact model— quality, believability, importance, and memorability—perform in other fear appeal studies including UV photo interventions. Moreover, a study using biometrics – such as skin conductance, heartrate, facial expression analysis, or EEG – might help to explicate the physiological impacts these variables make and unpack the relationship between message features and behavioral outcomes. In the current study, novelty was negatively correlated with physiological arousal, which suggests that participants’ low arousal is related to higher perception of novelty. Further investigation is required to understand if this finding is a spurious relationship, or if there is a meaningful causal relationship between these two important constructs? For example, a previous study demonstrated an 106 opposite pattern; high-novelty HIV/AIDS PSAs yielded greater self-reported arousal (Zhang et al., 2016). However, there was no significant differences in physiological arousal measured through skin conductance. Thus, future studies should measure both self-reported and physiological arousal with multiple forms of novel as well as traditional stimuli to explicate this relation. Personalized UV photos are a form of tailored message and perceived message relevance is one of the key mediators of tailored message (Jensen, King, Carcioppolo, & Davis, 2012; Kreuter & Wray, 2003). Thus, future research should study the mediating role of perceived message relevance in UV photo intervention. The existing measure of perceived message relevance (Jensen et al., 2012) might need some adaptation to be used with visual messages. In a longitudinal study, tailored interventions produced significantly greater sun protection behaviors compared to generic information. It would be interesting to study the impacts of PUV visuals in comparison with other forms of tailored information used by Manne and colleagues (2010) such as skin cancer risk and perceived benefit/barriers. Future studies could also compare the impacts of a personalized UV visual with personalized skin cancer risk results provided by the genetic testing of MC1R gene (Glanz et al., 2013; Wu et al., 2016). This study has important theoretical implications. Using physiological measures to track the emotional reactions of the participants is one of the meaningful contribution of this study. Complementing self-reported measures with physiological measures will enable retrieve information about participants’ conscious mind as well as unbiased physiological responses (Cacioppo et al., 2007). Thus, existing theoretical frameworks could be evaluated—potentially refined, if needed—by including physiological variables. 107 To that end, the current study used physiological arousal and valence in addition to the self-reported measures to evaluate postulates of the EPPM. Using physiological data to inform, evaluate, and advance the theoretical understanding of communication processes is an important next step. Thus, future studies should continue to integrate physiological measures to better understand and strengthen our theoretical frameworks. This study has several limitations. First, the sample size of the study is small (N = 108), especially considering that there are six experimental conditions, this limits the statistical power to detect small effects. This might be the explanation for not some of the null findings. Future studies should replicate the study with greater sample size to confirm the findings of this study. A second limitation is that the outcome measured was behavior expectations, thus it is unclear how long these changes would last and if these behavior expectations would lead to change in actual sun-safe behavior. Also, I have only studied the impacts of PUV visuals in the context of sun-safe behavior; other skin cancer prevention behaviors such as skin self-examination should be studied in future studies. Moreover, using spectrophotometry to track the objective measure of skin color (Mahler et al., 2013; Stock et al., 2010) in a longitudinal UV intervention study will be helpful in exploring the underlying mechanisms of effects. Third, only positive and negative valence were measured using facial expression analysis only. Further exploration of discrete emotions should be done in future studies by complementing physiological measure of emotions with self-reported items. A fourth limitation is that these analyses studied the danger control process portion of the EPPM model, but future UV intervention studies could test the full theoretical model of EPPM (including the fear control process). Researchers have also argued that fear has a less central role in the 108 EPPM, despite multiple studies depicting a positive role of fear in producing behavioral outcomes (Dillard, Li, Meczkowski, et al., 2016). Thus, a study evaluating a original theoretical model of EPPM and comparing it with the model where fear is more central might illuminate the confusion regarding fear-persuasion relationship. Despite the limitations, the current study adds to the literature on UV photo interventions, fear appeal, and physiological measures. The current analysis examined the value of personalized UV visuals compared to a stock UV, and a non-UV visual and demonstrated the potential of personalized as well as stock UV photo to promote positive behavior expectations through novelty. This provides a critical opportunity to synthesize the affective reactions produced when encountered with a fear-inducing stimulus. Table 4.1 Bivariate Correlations 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1 SUVvsPUV 2 NUVvsPUV ------- ---- 3 NUVvsSUV ---- ---- 4 Efficacy Stimuli .00 -.04 ----.04 ---- 5 Susceptibility .19 .22† .04 -.03 ---- 6 Severity .07 .23† .15 .00 .32** 7 Self Efficacy .05 .03 -.02 .12 -.05 ---.18† ---- 8 Response Efficacy -.02 .10 .12 .15 .30** .40** .41** 9 Fear .19 .29* .13 -.09 .07 .30** .18† .24* ---- .09 .05 -.03 -.04 -.03 .03 -.05 .01 -.09 .08 .10 .01 .03 -.17† -.03 .14 .02 .03 .04 .20 .30* .18 -.13 -.03 .00 .12 .11 .11 -.05 ---.06 ---- .06 .18 .10 .11 .10 .23* .37** .40** .19* -.07 .05 .12 .29 * .19 -.09 .09 .10 .21* .52** .26** .15 -.14 .01 .05 .56** .18 .11 -.08 .03 -.03 .19* .45** .18† .17† -.11 -.02 .11 .33** .61** 10 11 12 Physiological Arousal Negative Valence Positive Valence 15 16 17 ---- ---- ---- 15 Beh Exp_ Sunscreen Beh Exp_ Protective Items Beh Exp_ Seeking Shade 16 Novelty -.08 .24* .33** .03 .05 .21* .13 .30** .49** -.28** -.07 .07 .36** .27** .25** 17 Age -.01 -.10 -.09 -.07 -.19† .05 .20* -.13 -.02 -.11 .18† -.01 -.01 .22* .14 ----.13 Sex .20 † -.01 .19 .20* .14 .28** .08 .25** .26** -.15 .02 .03 .28** -.01 -.04 .17† .26** BRAT .16 .15 -.01 -.09 .38** .13 -.02 .17† .00 -.04 -.13 .12 .03 -.01 -.14 .00 -.13 13 14 18 19 18 ---------- ----.03 †p < .10, *p < .05, **p < .01 109 Table 4.2 Means and Standard Errors for Mediators and Outcomes by Visual and Efficacy Condition Visual conditions Efficacy conditions PUV SUV NUV No Efficacy Efficacy Susceptibility Severity Self Efficacy 5.23(.21) 4.75(.21) 4.66(.21) 4.91(.17) 4.85(.17) 5.99(.16) 5.87(.15) 5.58(.15) 5.81(.13) 5.82(.13) 4.80(.16) 4.71(.15) 4.73(.15) 4.64(.13) 4.85(.13) Response Efficacy Fear 5.96(.14) 6.00(.14) 5.80(.14) 5.80(.11) 6.05(.11) 4.17(.27)a 3.61(.26)ac 3.19(.26)bc 3.78(.21) 2.54(.21) Sunscreen Behavior Expectation 6.16(.22) 6.02(.22) 5.72(.22) 5.81(.18) 6.12(.18) Protective Items Behavior Expectation 5.72(.22)a 4.98(.22)b 5.21(.22)a 5.18(.18) 5.42(.18) Seeking Shade Behavior Expectation 5.06(.31) 4.41(.30) 4.68(.30) 4.66(.18) 4.77(.18) Novelty 5.07(.21)a 5.26(.21)a 4.41(.21)b 4.86(.17) 4.97(.17) Note. Means and standard errors (in parentheses). Means with different superscripts are significantly different, p < .10. 110 Table 4.3 Means and Standard Errors for Physiological Indicators (in time percent) by Visual and Efficacy Condition Physiological Arousal Negative Valence Positive Valence Visual conditions PUV 4.95(.68) 1.71(.78) 7.03(1.67)a SUV 4.33(.66) .90(1.63) 2.17(1.63)ac NUV 4.60(.67) .80(1.63) .11(1.63)bc Efficacy conditions No Efficacy 4.79(.55) .99(.63) 4.45(1.34) Efficacy 4.47(.55) 1.28(.63) 1.76(1.34) Note. Means and standard errors (in parentheses). Means with different superscripts are significantly different, p < .05. 111 112 Fear PUV vs. SUV .09(.29) Behavioral Expectation (Sunscreen) .14(.29) Note. Fear did not significantly mediate the relation between X and Y: effect = .06, Boot SE = .08, 95% Boot CI: -.0732, .2280. *p<.05 Fear PUV vs. NUV .26(.30) Behavioral Expectation (Sunscreen) .43(.29) Note. Fear did not significantly mediate the relation between X and Y: effect = .18, Boot SE = .11, 95% Boot CI: -.0008, .4124. *p<.05 Fear PUV vs. SUV .69(.30)* Behavioral Expectation (Protective items) .75(.31)* Note. Fear did not significantly mediate the relation between X and Y: effect = .06, Boot SE = .08, 95% Boot CI: -.0239, .2671. *p<.05 Figure 4.1 Simple mediation models (Fear as mediator) 113 Fear PUV vs. NUV .40(.32) Behavioral Expectation (Protective items) .50(.31) Note. Fear did not significantly mediate the relation between X and Y: effect = .10, Boot SE = .09, 95% Boot CI: -.0518, .3007. *p<.05 Fear PUV vs. SUV .53(.44) Behavioral Expectation (Seeking Shade) .65(.43) Note. Fear did not significantly mediate the relation between X and Y: effect = .13, Boot SE = .1425, 95% Boot CI: -.0561, .4928. *p<.05 Fear PUV vs. NUV .20(.41) Behavioral Expectation (Seeking Shade) .38(.40) Note. Fear did not significantly mediate the relation between X and Y: effect = .18, Boot SE = .16, 95% Boot CI: -.0792, .5538. *p<.05 Figure 4.1 contd… 114 Physiological Arousal PUV vs. SUV .21(.28) Behavioral Expectation (Sunscreen) .14(.29) Note. Physiological arousal did not significantly mediate the relation between X and Y: effect = -.06, Boot SE = .10, 95% Boot CI: -.2783, .1016. *p<.05 Physiological Arousal PUV vs. NUV .43(.29) Behavioral Expectation (Sunscreen) .43(.29) Note. Physiological arousal did not significantly mediate the relation between X and Y: effect = .00, Boot SE = .04, 95% Boot CI: -.1174, .0669. *p<.05 Physiological Arousal PUV vs. SUV .80(.29)** Behavioral Expectation (Protective items) .75(.30)* Note. Physiological arousal did not significantly mediate the relation between X and Y: effect = -.05, Boot SE = .08, 95% Boot CI: -.2244, .0994. *p<.05 **p<.01 Figure 4.2 Simple mediation models (Physiological Arousal as mediator) 115 Physiological Arousal PUV vs. NUV .52(.31)† Behavioral Expectation (Protective items) .50(.31) Note. Physiological arousal did not significantly mediate the relation between X and Y: effect = -.02, Boot SE = .06, 95% Boot CI: -.1829, .0867. *p<.05 †p < .10 Physiological Arousal PUV vs. SUV .71(.43) Behavioral Expectation (Seeking Shade) .65(.43) Note. Physiological arousal did not significantly mediate the relation between X and Y: effect = -.06, Boot SE = .10, 95% Boot CI: -.2915, .1106. *p<.05 Physiological Arousal PUV vs. NUV .41(.39) Behavioral Expectation (Seeking Shade) .38(.40) Note. Physiological arousal did not significantly mediate the relation between X and Y: effect = -.03, Boot SE = .12, 95% Boot CI: -.3729, .0920. *p<.05 †p < .10 Figure 4.2 contd… 116 Positive Valence PUV vs. SUV .31(.34) Behavioral Expectation (Sunscreen) .33(.33) Note. Positive Valence did not significantly mediate the relation between X and Y: effect = .02, Boot SE = .06, 95% Boot CI: -.1083, .1288. Positive Valence PUV vs. NUV .50(.35) Behavioral Expectation (Sunscreen) .48(.33) Note. Positive Valence did not significantly mediate the relation between X and Y: effect = -.02, Boot SE = .09, 95% Boot CI: -.1455, .2378. *p<.05 Positive Valence PUV vs. SUV .46(.29) Behavioral Expectation (Protective items) .45(.28) Note. Positive Valence did not significantly mediate the relation between X and Y: effect = -.08, Boot SE = .08, 95% Boot CI: -.2103, .0772. *p<.05 Figure 4.3 Simple mediation models (Positive Valence as mediator) 117 Positive Valence PUV vs. NUV .42(.35) Behavioral Expectation (Protective items) .32(.33) Note. Positive valence did not significantly mediate the relation between X and Y: effect = -.09, Boot SE = .10, 95% Boot CI: -.2839, .0839. *p<.05 Positive Valence PUV vs. SUV .31(.38) Behavioral Expectation (Seeking Shade) .37(.37) Note. Positive valence did not significantly mediate the relation between X and Y: effect = .05, Boot SE = .05, 95% Boot CI: -.0436, .1491. *p<.05 Positive Valence PUV vs. NUV -.10(.42) Behavioral Expectation (Seeking Shade) -.07(.40) Note. Positive valence did not significantly mediate the relation between X and Y: effect = .03, Boot SE = .08, 95% Boot CI: -.0724, .2354. *p<.05 **p<.01 Figure 4.3 contd… 118 Novelty PUV vs. NUV .20(.27) Behavioral Expectation (Sunscreen) .43(.29) Note. Novelty significantly mediated the relation between X and Y: effect = .24, Boot SE = .12, 95% Boot CI: .0073, .3939. *p<.05 Novelty PUV vs. SUV .21(.28) Behavioral Expectation (Sunscreen) .14(.29) Note. Novelty did not significantly mediate the relation between X and Y: effect = -.07, Boot SE = .11, 95% Boot CI: -.3009, .1439. *p<.05 Novelty SUV vs. NUV .00(.35) Behavioral Expectation (Sunscreen) .29(.35) Note Novelty significantly mediated the relation between X and Y: effect = .29, Boot SE = .15, 95% Boot CI: .0103, .4161. *p<.05 Figure 4.4 Simple mediation models (Novelty as a mediator) 119 Novelty PUV vs. NUV .28(.30) Behavioral Expectation (Protective items) .50(.31) Note. Novelty significantly mediated the relation between X and Y: effect = .22, Boot SE = .12, 95% Boot CI: .0099, .4880. *p<.05 Novelty PUV vs. SUV .82(.29) * Behavioral Expectation (Protective items) .75(.30)* Note. Novelty did not significantly mediate the relation between X and Y: effect = -.07, Boot SE = .11, 95% Boot CI: -.3068, .1258. *p<.05 Novelty SUV vs. NUV .46(.34) Behavioral Expectation (Protective items) -.25(.33) Note. Novelty did not significantly mediate the relation between X and Y: effect = .21, Boot SE = .13, 95% Boot CI: -.0008, .5194. *p<.05 †p<.10 Figure 4.4 contd… 120 Novelty PUV vs. NUV .10(.39) Behavioral Expectation (Seeking Shade) .38(.40) Note. Novelty significantly mediated the relation between X and Y: effect = .28, Boot SE = .17, 95% Boot CI: .0118, .6512. *p<.05 Novelty PUV vs. SUV .74(.42) † Behavioral Expectation (Seeking Shade) .65(.43) Note. Novelty did not significantly mediate the relation between X and Y: effect = .09, Boot SE = .15, 95% Boot CI: -.4382, .1706. *p<.05 †p<.10 Novelty SUV vs. NUV -.50(.45) Behavioral Expectation (Seeking Shade) -.28(.43) Note. Novelty did not significantly mediate the relation between X and Y: effect = .22, Boot SE = .17, 95% Boot CI: -.0456, .6226. *p<.05 Figure 4.4 contd… 121 4.7 Supplemental Materials 4.7.1 Text Prompt Preceding the Visual Conditions 122 4.7.2 SUV Visual 123 4.7.3 NUV Visual 124 4.7.4 Efficacy Stimuli CHAPTER 5 CONCLUDING REMARKS The current dissertation focused on visuals as a means of communicating health information explicitly related to cancer. More specifically, this dissertation launched three studies to initiate a research program that investigates the persuasive impact of visual message features on knowledge, attitudes, and behavior in the context of cancer prevention and control. To that end, the first study was focused on all cancer whereas the second and third studies focused on skin cancer. The second and third studies evaluated the impact of ultraviolet photos by comparing them with other forms of visuals (such as visuals showing people being exposed to UV light, sunburn, photo-aging, and mole removal scars) used to promote sun-safe behaviors and prevent skin cancer. These two studies evaluated the key postulates of the extended parallel process model and investigated the role of fear in the model. In addition to the theoretical contribution, the dissertation also exhibited multiple methodological perspectives as all the three studies employ a different methodological approach. To that end, the first study was a meta-analysis, second was a large sample web-based study, and the third was a lab-based experiment using physiological measures in addition to the self-reported survey items. Collectively, these three studies—presented in chapters two, three, and four respectively—made several theoretical and empirical 126 contributions to health communication research. This chapter provides the summary of each of these studies, their contributions, and future steps of the research program. The second chapter reported a meta-analysis that examined the efficacy of visual interventions in communicating cancer risk. The goal of the meta-analysis was to quantify the impact of visuals in cancer risk communication and identify critical future directions for research in this area. The meta-analysis summarized the findings of six studies, eight effect sizes (N = 5,832) and found a small but positive impact of including visuals in communicating cancer risk. Chapter two offered suggestions regarding the use of appropriate visuals in health campaigns and suggested researchers to focus future work on explicating the effects of several other components of the visuals—for example, frequency, picture quality, density, and light spectrum—which were not included in this meta-analysis but are important to broaden our understanding of visuals. The next logical step would be summarizing the impact of visuals in all health contexts—not just related to cancer. Moreover, extending the findings of this meta-analyses by including behavioral outcomes would be a valuable contribution to the literature. The third chapter reported a large sample web-based experimental study with a sophisticated study design: a 5 (visual conditions) × 3 (replications) × 4 (efficacy conditions) between-participants experiment. The visuals, data, and approach of this study serve as a resource to inform and support future studies on visual communication, UV visuals, skin cancer, and fear appeals. This study was in line with the recommendations of a recent meta-analysis of fear appeal research that advocated for more studies directly testing the mediating role of fear and not just using it as a manipulation check variable (Tannenbaum et al., 2015). Chapter three demonstrated that 127 stock UV images have a positive influence on sun-safe behavior expectations via fear. This is interesting given that the fear appeal model—EPPM—considers fear as an aversive affective state and may not fully recognize its potential to have a positive role in persuasion. This study engages Tannenbaum and colleagues’ call to action, however, even more work is needed to unpack the persuasive impact of fear appeals and the role that other emotions play in the process. Chapter 4 provided a lab-based study which engaged questions generated by the study in chapter three and moved forward in two important ways: (1) adding a personalized UV (PUV) photo condition and (2) utilizing physiological measures (i.e., facial expression and skin conductance). This study contributes to existing research by comparing personalized and stock UV photos, a move that provides data examining the relative value of providing participants with a photo of their own face rather than a stock image. The results suggest that both personalized and stock UV images have potential utility in skin cancer prevention campaigns. For example, when compared with the NUV condition, the PUV condition produced greater self-reported fear but also produced greater positive valence as seen by tracking physiological responses. Additionally, the results demonstrated the importance of message novelty. Novelty significantly mediated the relationship between exposure to PUV vs. NUV (and SUV vs. NUV) skin damage visuals and all three types of behavior expectations (i.e., sunscreen, protective items, and seeking shade). 128 5.1 Contributions Overall, this dissertation made contributions to visual health communication in the context of cancer. Each of the chapters put forward researchers and practitioners understanding of the persuasive impact of visuals. The contributions of the dissertation can be organized into three larger areas of communication. First, the program helped explicate key features of visuals and evaluated their impacts. Second, it advanced our understanding of fear and other affective variables—including physiological measures— as a mediator in communication efforts. Third, it provided an evidence base for cancer communication and control, notably in the context of skin cancer. 5.1.1 Visual Features and Impact The first study—a meta-analysis—demonstrated the utility of visuals in communicating cancer risk, whereas the second and third study explicated the persuasive impacts and features of visuals, mainly based on two critical dimensions: light spectrum and personalization or tailoring. Additionally, study three identified the similarities and differences between personalized and stock UV photos by comparing them against a mole removal visual. The second study utilized multiple visual categories with replications (e.g., Jackson & Jacobs, 1983), which is rarely encountered in communication research (Jensen, 2008). The study examined the value of stock UV photos compared to four other visual types (i.e., UV vs. all). Morever, other researchers could utilize the data to answer alternative research questions about the impact of visual categories. 129 5.1.2 Affective Constructs as a Mediator Chapters three and four added to the scholarly conversation about investigating the mediating role of fear and other discrete emotions. The third study, in particular, used physiological measures to track the emotional reactions of the participants and tested them as potential mediators. Physiological measures provide a critical opportunity to track the affective reactions produced when encountered with a fear-inducing stimulus. Using physiological arousal and valence as a mediator in the existing theoretical framework is an important contribution of this study as this advances the theoretical understanding of communication processes. These chapters collectively showed that a mixture of emotions—rather than just fear—might interplay to promote behavioral outcomes, as reported in several other studies (for a review, see Dillard & Nabi, 2006). Future studies should refine the existing self-reported measures of discrete emotions (Dillard & Shen, 2006) and triangulate them with physiological measures to explore the mediating role played by the affective constructs. 5.1.3 Evidence Base for Cancer Communication and Control The findings of this dissertation have some practical implications in cancer communication and control efforts. The dissertation found critical differences between personalized UV photos, stock UV photos, and mole removal photos. Other important findings from the program relate to the importance of novelty in messages. When messages are perceived as novel, they are more effective in influencing behavioral 130 outcomes. This has important practical implications for cancer prevention efforts, notably skin cancer. For example, including a novel message feature such as UV photos might produce greater sun-safe behavioral outcomes. Thus, skin cancer prevention materials and social media pages of organizations working to promote sun-safe behaviors should include UV photos showing varying level of damages. However, there are some unanswered questions about efficacy beliefs. Thus, future studies should continue to assess the impacts of efficacious messages to unpack the relationship between selfefficacy, fear, and behavioral outcomes. As the pursuit of research in this area continues to move forward, the findings of these studies will continue to provide evidence for the development of interventions and campaigns in the context of cancer communication and control. 5.2 Moving Forward Health communication practitioners often create visuals to encourage or motivate behavior change (Houts et al., 2006; McWhirter & Hoffman-Goetz, 2014). One strategy is to use visuals that directly or indirectly communicate risk. Such visuals are often labeled risk or fear appeals as communicators assume they exert influence by triggering one or both responses. Researchers have made similar assumptions because risk and fear are conceptualized as explanatory variables in theoretical frameworks (e.g., the EPPM; Witte, 1992). Yet, cognitive response to message features can be unexpected and nonlinear (see, e.g., O’Keefe & Jensen, 2006). It seems logical to postulate that risk and fear appeals exert influence via risk and fear, but cognitive response does not always mirror message features. 131 The current research suggests that there is considerably more work to be done explicating the features and pathways of visuals that seem to target risk and fear. On the one hand, there is evidence that risk visuals do increase perceived risk (chapter two) and that fear appeals do increase perceived fear (chapter three). On the other hand, fear appeals do not always trigger increased risk (chapters three and four) and they can generate both fear and positive affect (chapter four). Collectively, these findings raise questions about the relationship between risk and fear, what features drive fear if risk remains unchanged, how best to label/categorize such visuals, and how positive affect and fear can coexist. 5.2.1 Future Studies The meta-analysis reported in the second chapter demonstrated the positive impact of including visuals in cancer risk communication. However, several questions in this area remain unanswered. First, there is a need for more research comparing text stimuli with text and visual stimuli in communicating risk information; not only to advance our understanding of the benefit of visuals but also to increase power of the aforementioned meta-analysis. Concerning the latter, additional studies might have relatively substantial impacts on the results such as revealing significant moderators including type of visuals, participant characteristics, study design, and type of cancer. A logical extension of the meta-analysis is to include behavioral variables as an outcome to determine if the increased perceptions of risk leads to behavioral actions. Further, it would be interesting to summarize the impacts of visuals on accuracy of risk perception, not just magnitude of risk perception. 132 The key finding of the visual web-based experimental study reported in the third chapter is that UV visual interventions compared to other visual interventions produce significantly greater sun-safe behavior expectations via fear. In contrary to the proposition of the EPPM, yet in line with the discussions raised by others (e.g., Dillard, Li, & Huang, 2016), fear was positively related to behavior change. That raises questions about which theoretical framework, if any, would be suitable to explain these positive impacts. Future studies should consider if reevaluating the postulates of the EPPM could resolve this uncertainty or if explicating other existing fear appeal theories such as drive theory (Hovland et al., 1953)—which postulates a direct relation of fear with behavioral outcome—might be more reasonable. In fact, a related question raised by this study is about the mechanisms that fear appeal visuals utilize to create a behavioral impact. Do some fear inducing visuals impact through the channel of perceived susceptibility and severity—as postulated by the EPPM—and other visuals impact via fear, as seen in the third chapter of this dissertation. If so, what is the distinguishing characteristic that determines the mechanism? Alternatively, what does that say about the postulates of the EPPM? A series of studies explicating the constructs and features of fear, fear appeal visuals, and their relation with persuasive communication is required to answer this deeper issue. Another possible avenue for future studies could be examining the impact of UV photo interventions using the tri-partite model of risk perceptions (TRIRISK; Ferrer, Klein, Persoskie, Avishai-Yitshak, & Sheeran, 2016). The model identifies three types of risk: deliberative, affective, and experiential risk. Deliberative risk is based on logical cognitive thinking such as, the possibility of contracting a disease. Affective risk 133 perceptions, on the other hand, include emotional valence (both negative and positive) and arousal in response to the threat. Experiential risk is a “gut-level assessment of vulnerability” (p. 654). Ferrer and colleagues (2016) demonstrated that these three risk perceptions function as unique predictors to influence behavior intentions. The studies in this dissertation have revealed that UV photo interventions primarily function through fear, and not susceptibility or severity, thus it might be interesting to evaluate if we see similar patterns with these three types of risk. In addition, the third chapter exhibited promising impacts of stock UV visuals in promoting sun-safe behaviors. To extend this finding, future studies could explore the impact of other forms of stock UV photos. For example, providing comparative visuals showing the face in both natural and UV light might influence sun-safe behaviors differently. Another approach could be using a series of stock UV photos depicting real (or simulated) damages over time. A different strategy could be showing multiple stock UV photos of an assortment of people of the same age, but varying skin damage. Another novel way to communicate difference in skin damage caused by sun exposure over time is to provide a UV photo of a child—which should have significantly less damage—and pair it with UV photo of an adult. This visual will magnify contrast in ways that may maximize the impact of the image. The impacts of these various combinations of stock UV photos could be compared with the impacts made by a single stock UV photo to understand the added utility of providing comparison stock UV photos. This research could be supplemented by a dyadic research—notably a parent-child dyad. To this end, seeing skin damage in a child’s UV image might produce a significant change in parental attitudes and behavior related to skin cancer prevention. 134 The lab study reported in the fourth chapter of this dissertation revealed that exposure to a personalized UV visual produces significantly greater fear when compared to a non-UV visual (i.e., mole removal visual), but not when compared with a stock UV visual. However, future research with a larger sample size should investigate the difference between a personalized and a stock UV photo as the current study had limited statistical power to identify moderators and small (but expected) effects. Having a larger sample would also provide sufficient statistical power to evaluate the mediating role of fear and other affective variables, which were also not significant in the lab study. Emotional reactions have been found to be effective in influencing sun-safe behaviors (Mahler, 2014, 2018). In a recent study, Mahler (2018) discussed the importance of including physiological measures to enhance understanding of affective processes. Accordingly, the lab study included physiological measures of arousal— measured using skin conductance—and valence—measured by facial expression analysis. The results demonstrated that positive valence was significantly greater when participants were exposed to their personalized UV visual as compared to a non-UV visual. An important limitation of the lab study is that self-reported measures of discrete emotions were not included. Future research should expand the type and number of discrete emotions examined as mediators. Fear appeals may exert impact via fear and other emotions (Dillard & Nabi, 2006; Leshner, Bolls, & Wise, 2011; Nabi & Myrick, 2018); a reality that requires researchers to investigate a wider range of affect as mediational pathways. Researchers could pursue this goal by including a self-report measure that examines multiple discrete emotions (Dillard & Shen, 2006) and/or using physiological measures to assess a wider range of affective constructs. To this end, a follow-up lab 135 study with the same three visual experimental conditions (PUV, SUV, and NUV) as the lab study of this dissertation has been designed to include the self-reported measures of discrete emotions and physiological measures of arousal and valence. Yet, lab studies of this type often have small samples because data collection is time consuming; therefore, significant progress on this front will likely take considerable time and the efforts of multiple research teams before generalizable results can be identified. Advancing measurement of affective states is another key research goal. For example, it is important to explicate the feelings of participants who report fear and positive affect after exposure to personalized UV photos. How might this perception best be characterized? One way to investigate this would be to further examine the impacts of personalized and stock UV photos in a qualitative study. Qualitative research would provide an opportunity for participants to describe their reactions in ways that may suggest new avenues and directions for affective measurement. In addition, it would be valuable to record the physiological responses of the participants as they are reflecting on their UV photos. This would reveal the thinking processes, accompanying feelings, and underlying physiological responses that participants undergo during and after exposure to the personalized and stock UV photos. Another avenue of research is to conceptualize personalized UV photos as tailored messages. Tailored messages are known to create an impact through perceived message relevance (Jensen et al., 2012; Kreuter & Wray, 2003). However, the measures of perceived message relevance (Jensen et al., 2012) might need some adaptation to be used with visual stimuli; thus, future studies using personalized UV photos should adapt existing perceived relevance measures and evaluate the role of perceived relevance in UV 136 photo interventions. Finally, there are also some unresolved questions about the role of efficacy beliefs. The results of the web-based experiment questioned the moderating role of efficacy messages by demonstrating that the indirect effect through fear was significant for less efficacious conditions (i.e., control and text only). Contrary to the theoretical assumptions and proposed hypothesis, efficacious visual messages appeared to nullify fear, and in the absence of the fear, the indirect effect was no longer significant. Replication of this finding is a logical next step. Following that, researchers should examine whether different features of efficacy appeals impact their effect. For instance, Carcioppolo and colleagues (2013) postulated that message dose could moderate the impact of efficacy appeals. Future studies should examine different dosages of efficacy appeals, notably in relation to baseline efficacy, and examine whether fear nullification manifests at a particular threshold. 5.3 Conclusion Overall, the dissertation makes several contributions to the study of visual message features, fear appeals, and persuasion. First, this dissertation demonstrates the utility of visuals in public health campaigns communicating cancer risks. The studies in the dissertation demonstrated that both personalized and stock UV visuals could be conceptualized as fear appeals. More specifically, it reveals the utility of a stock UV photo in promoting sun-safe behaviors through fear. In addition, the dissertation revealed some important differences among a personalized UV photo, a stock UV photo, and a mole removal photo. The studies also demonstrated the mediating novelty in promoting 137 sun-safe behaviors such that when participants perceived UV photos as novel, they were more likely to report positive behavior expectations. Taken together, these three studies thread together a unique program of research in visual interventions and cancer communication. The findings of the three studies provide an evidence base for interventions as well as public health campaigns. REFERENCES Allen, M. (2009). Meta-Analysis. Communication Monographs, 76(4), 398–407. https://doi.org/10.1080/03637750903310386 An, D. (2007). Advertising visuals in global brands’ local websites: A six-country comparison. International Journal of Advertising, 26(3), 303. https://doi.org/10.1080/02650487.2007.11073016 Ancker, J. S., Senathirajah, Y., Kukafka, R., & Starren, J. B. (2006). Design features of graphs in health risk communication: A systematic review. Journal of the American Medical Informatics Association, 13(6), 608–618. https://doi.org/10.1197/jamia.M2115 Armitage, C. J., Norman, P., Alganem, S., & Conner, M. (2015). Expectations are more predictive of behavior than behavioral intentions: Evidence from two prospective studies. Annals of Behavioral Medicine, 49(2), 239–246. https://doi.org/10.1007/s12160-014-9653-4 Aspinwall, L. G., Leaf, S. L., Kohlmann, W., Dola, E. R., & Leachman, S. A. (2009). Patterns of photoprotection following CDKN2A/p16 genetic test reporting and counseling. Journal of the American Academy of Dermatology, 60(5), 745–757. https://doi.org/10.1016/j.jaad.2008.12.034 Aspinwall, L. G., Taber, J. M., Kohlmann, W., Leaf, S. L., & Leachman, S. A. (2014). Unaffected family members report improvements in daily routine sun protection 2 years following melanoma genetic testing. Genetics in Medicine, 16(11), 846– 853. https://doi.org/10.1038/gim.2014.37 Aue, T., Hoeppli, M.-E., & Piguet, C. (2012). The sensitivity of physiological measures to phobic and nonphobic fear intensity. Journal of Psychophysiology, 26(4), 154– 167. https://doi.org/10.1027/0269-8803/a000078 Baker, D. W. (2007). Health literacy and mortality among elderly persons. Archives of Internal Medicine, 167(14), 1503. https://doi.org/10.1001/archinte.167.14.1503 Bandura, A. (2004). Health promotion by social cognitive means. Health Education & Behavior, 31(2), 143–164. https://doi.org/10.1177/1090198104263660 Beasley, T. M., & Kittel, B. S. (1997). Factors that influence health risk behaviors among tanning salon patrons. Evaluation & The Health Professions, 20(4), 371–388. https://doi.org/10.1177/016327879702000401 139 Birmingham, W. C., Hung, M., Boonyasiriwat, W., Kohlmann, W., Walters, S. T., Burt, R. W., … Kinney, A. Y. (2015). Effectiveness of the extended parallel process model in promoting colorectal cancer screening. Psycho-Oncology, 24(10), 1265– 1278. https://doi.org/10.1002/pon.3899 Borenstein, M., Hedges, L. H., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to Meta-Analysis. West Sussex, UK: John Wiley & Sons Ltd. Bradley, M. M., & Lang, P. J. (1988). Measuring emotion: Behavior, feeling and physiology. In R. Lane & L. Nadel (Eds.), Cognitive neuroscience of emotion (pp. 261–302). New York: Oxford University Press. Brehm, J. W. (1966). A theory of psychological reactance. Oxford, England: Academic Press. Brehm, S. S., & Brehm, J. W. (1981). Psychological reactance: A theory of freedom and control. Oxford, England: Academic Press. Cacioppo, J. T., Tassinary, L. G., & Berntson, G. (2007). Psychophysiological science: Interdisciplinary approaches to classic questions about the mind. In J. T. Cacioppo, L. G. Tassinary, & G. Berntson (Eds.), Handbook of psychophysiology (3rd ed., pp. 1–16). Cambridge, UK: Cambridge University Press. Carcioppolo, N., Jensen, J. D., Wilson, S. R., Collins, W. B., Carrion, M., & Linnemeier, G. (2013). Examining HPV threat-to-efficacy ratios in the extended parallel process model. Health Communication, 28(1), 20–28. Retrieved from ufh. Carrera, P., Muñoz, D., & Caballero, A. (2010). Mixed emotional appeals in emotional and danger control processes. Health Communication, 25(8), 726–736. https://doi.org/10.1080/10410236.2010.521914 Cauberghe, V., De Pelsmacker, P., Janssens, W., & Dens, N. (2009). Fear, threat and efficacy in threat appeals: Message involvement as a key mediator to message acceptance. Accident Analysis & Prevention, 41(2), 276–285. https://doi.org/10.1016/j.aap.2008.11.006 Chen, Y., & Yang, Z. J. (2015). Message formats, numeracy, risk perceptions of alcoholattributable cancer, and intentions for binge drinking among college students. Journal of Drug Education, 45(1), 37–55. https://doi.org/10.1177/0047237915604062 Cohen, J. (1988). Statistical power analysis for the social sciences. Hillsdale, NJ: Erlbaum. Cortina, J. M. (2003). Apples and oranges (and pears, oh my!): The search for moderators in meta-analysis. Organizational Research Methods, 6(4), 415–439. https://doi.org/10.1177/1094428103257358 140 Couper, M. P. (2000). Web surveys: A review of issues and approaches. The Public Opinion Quarterly, 64(4), 464–494. Cummins, R. G., Keene, J. R., & Nutting, B. H. (2012). The impact of subjective camera in sports on arousal and enjoyment. Mass Communication & Society, 15(1), 74– 97. Retrieved from ufh. De Meulenaer, S., De Pelsmacker, P., & Dens, N. (2015). Have no fear: How individuals differing in uncertainty avoidance, anxiety, and chance belief process health risk messages. Journal of Advertising, 44(2), 114–125. https://doi.org/10.1080/00913367.2015.1018465 Demierre, M.-F., Maguire-Eisen, M., O’Connell, N., Sorenson, K., Berger, J., Williams, C., & Cabral, H. (2009). A sun protection community intervention in Quincy middle schools: Insights from the use of ultraviolet photography and its impact on sunburn. Journal of the Dermatology Nurses’ Association, 1(2), 111–118. Dijksterhuis, A. (2004). Think different: The merits of unconscious thought in preference development and decision making. Journal of Personality and Social Psychology, 87(5), 586–598. https://doi.org/10.1037/0022-3514.87.5.586 Dillard, J. P., Li, R., & Huang, Y. (2016). Threat appeals: The fear–persuasion relationship is linear and curvilinear. Health Communication, 1–10. https://doi.org/10.1080/10410236.2016.1220345 Dillard, J. P., Li, R., Meczkowski, E., Yang, C., & Shen, L. (2016). Fear responses to threat appeals: Funtional form, methodological considerations, and correspondence between static and dynamic data. Communication Research, 44(7), 1–22. https://doi.org/10.1177/0093650216631097 Dillard, J. P., & Nabi, R. L. (2006). The persuasive influence of emotion in cancer prevention and detection messages. Journal of Communication, 56(suppl_1), S123–S139. https://doi.org/10.1111/j.1460-2466.2006.00286.x Dillard, J. P., Plotnick, C., Godbold, L. C., Freimuth, V. S., & Edgar, T. (1996). The multiple affective outcomes of AIDS PSAs. Communication Research, 23(1), 44– 72. https://doi.org/10.1177/009365096023001002 Dillard, J. P., & Shen, L. (2005). On the nature of reactance and its role in persuasive health communication. Communication Monographs, 72(2), 144–168. https://doi.org/10.1080/03637750500111815 Dillard, J. P., & Shen, L. (2006). Self-report measures of discrete emotions. In R. A. Reynolds & J. D. Baker (Eds.), Handbook of research on electronic surveys and measurements (pp. 330–333). https://doi.org/10.4018/978-1-59140-792-8.ch044 141 Dillon, K. H., Allan, N. P., Cougle, J. R., & Fincham, F. D. (2015). Measuring hostile interpretation bias: The WSAP hostility scale. Assessment, 23(6), 707–719. https://doi.org/10.1177/1073191115599052 Emmons, K. M., Geller, A. C., Puleo, E., Savadatti, S. S., Hu, S. W., Gorham, S., & Werchniak, A. E. (2011). Skin cancer education and early detection at the beach: A randomized trial of dermatologist examination and biometric feedback. Journal of the American Academy of Dermatology, 64(2), 282–289. https://doi.org/10.1016/j.jaad.2010.01.040 Erdfelder, E., Faul, F., & Buchner, A. (1996). GPOWER: A general power analysis program. Behavior Research Methods, Instruments, & Computers, 28(1), 1–11. https://doi.org/10.3758/BF03203630 Fabrizi, G., Pagliarello, C., & Massi, G. (2008). A simple and inexpensive method for performing ultraviolet photography in your dermatology practice. Dermatologic Surgery, 34(8), 1093. Fagerlin, A., Zikmund-Fisher, B., Ubel, P. A., Jankovic, A., Derry, H. A., & Dylan, S. M. (2007). Measuring numeracy without a math test: Development of the subjective numeracy scale. Medical Decision Making : An International Journal of the Society for Medical Decision Making, 27(5), 672–680. Fazio, R. H., & Olson, M. A. (2003). Implicit measures in social cognition research: Their meaning and use. Annual Review of Psychology, 54(1), 297–327. https://doi.org/10.1146/annurev.psych.54.101601.145225 Ferrer, R. A., Klein, W. M. P., Persoskie, A., Avishai-Yitshak, A., & Sheeran, P. (2016). The tripartite model of risk perception (TRIRISK): Distinguishing deliberative, affective, and experiential components of perceived risk. Annals of Behavioral Medicine, 50(5), 653–663. https://doi.org/10.1007/s12160-016-9790-z Finucane, M. L., Slovic, P., Mertz, C. k., Flynn, J., & Satterfield, T. A. (2000). Gender, race, and perceived risk: The “white male” effect. Health, Risk & Society, 2(2), 159–172. https://doi.org/10.1080/13698570050077390 Fischhoff, B. (1999). Why (cancer) risk communication can be hard. Journal of the National Cancer Institute, 1999(25), 7. Fulton, J. E. (1997). Utilizing the ultraviolet (UV detect) camera to enhance the appearance of photodamage and other skin conditions. Dermatologic Surgery, 23(3), 163–169. https://doi.org/10.1111/j.1524-4725.1997.tb00013.x Gamble, R. G., Asdigian, N. L., Aalborg, J., Gonzalez, V., Box, N. F., Huff, L. S., … Dellavalle, R. P. (2012). Sun damage in ultraviolet photographs correlates with phenotypic melanoma risk factors in 12-year-old children. Journal of the American Academy of Dermatology, 67(4), 587–597. https://doi.org/10.1016/j.jaad.2011.11.922 142 Gandini, S., Sera, F., Cattaruzza, M. S., Pasquini, P., Abeni, D., Boyle, P., & Melchi, C. F. (2005). Meta-analysis of risk factors for cutaneous melanoma: I. Common and atypical naevi. European Journal of Cancer, 41(1), 28–44. https://doi.org/10.1016/j.ejca.2004.10.015 Gandini, S., Sera, F., Cattaruzza, M. S., Pasquini, P., Picconi, O., Boyle, P., & Melchi, C. F. (2005). Meta-analysis of risk factors for cutaneous melanoma: II. Sun exposure. European Journal of Cancer, 41(1), 45–60. https://doi.org/10.1016/j.ejca.2004.10.016 Gandini, S., Sera, F., Cattaruzza, M. S., Pasquini, P., Zanetti, R., Masini, C., … Melchi, C. F. (2005). Meta-analysis of risk factors for cutaneous melanoma: III. Family history, actinic damage and phenotypic factors. European Journal of Cancer, 41(14), 2040–2059. https://doi.org/10.1016/j.ejca.2005.03.034 Garcia-Retamero, R., & Galesic, M. (2010). Who proficts from visual aids: Overcoming challenges in people’s understanding of risks. Social Science & Medicine, 70(7), 1019–1025. https://doi.org/10.1016/j.socscimed.2009.11.031 Gerrard, M., Gibbons, F. X., Houlihan, A. E., Stock, M. L., & Pomery, E. A. (2008). A dual-process approach to health risk decision making: The prototype willingness model. Developmental Review, 28(1), 29–61. https://doi.org/10.1016/j.dr.2007.10.001 Gibbons, F. X., Gerrard, M., Lane, D. J., Mahler, H. I. M., & Kulik, J. A. (2005). Using UV photography to reduce use of tanning booths: A test of cognitive mediation. Health Psychology, 24(4), 358–363. https://doi.org/10.1037/0278-6133.24.4.358 Gigerenzer, G. (2007). Gut feelings: The intelligence of the unconscious. London, England: Penguin. Glanz, Karen, Schoenfeld, E., Weinstock, M. A., Layi, G., Kidd, J., & Shigaki, D. M. (2003). Development and reliability of a brief skin cancer risk assessment tool. Cancer Detection and Prevention, 27(4), 311–315. https://doi.org/10.1016/S0361090X(03)00094-1 Glanz, Karen, Volpicelli, K., Kanetsky, P. A., Ming, M. E., Schuchter, L. M., Jepson, C., … Armstrong, K. (2013). Melanoma genetic testing, counseling, and adherence to skin cancer prevention and detection behaviors. Cancer Epidemiology Biomarkers & Prevention, 22(4), 607. https://doi.org/10.1158/1055-9965.EPI-12-1174 Green, E., Mitchell, W., & Bunton, R. (2000). Contextualizing risk and danger: An analysis of young people’s perceptions of risk. Journal of Youth Studies, 3(2), 109–126. https://doi.org/10.1080/713684369 Guy Jr, G. P., Thomas, C. C., Thompson, T., Watson, M., Massetti, G. M., & Richardson, L. C. (2015). Vital signs: Melanoma incidence and mortality trends and 143 projections—United States, 1982–2030. Morbidity and Mortality Weekly Report, 64(21), 591–596. Hale, J. L., Lemieux, R., & Mongeau, P. A. (1995). Cognitive processing of fear-arousing message content. Communication Research, 22(4), 459–474. https://doi.org/10.1177/009365095022004004 Han, P. K. J., Klein, W. M. P., Killam, B., Lehman, T., Massett, H., & Freedman, A. N. (2012). Representing randomness in the communication of individualized cancer risk estimates: Effects on cancer risk perceptions, worry, and subjective uncertainty about risk. Patient Education and Counseling, 86(1), 106–113. https://doi.org/10.1016/j.pec.2011.01.033 Han, P. K. J., Klein, W. M. P., Lehman, T., Killam, B., Massett, H., & Freedman, A. N. (2011). Communication of uncertainty regarding individualized cancer risk estimates: Effects and influential factors. Medical Decision Making, 31(2), 354– 366. https://doi.org/10.1177/0272989X10371830 Hawley, S. T., Zikmund-Fisher, B., Ubel, P., Jancovic, A., Lucas, T., & Fagerlin, A. (2008). The impact of the format of graphical presentation on health-related knowledge and treatment choices. Patient Education and Counseling, 73(3), 448– 455. https://doi.org/10.1016/j.pec.2008.07.023 Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York: Guilford Press. Hayes, A. F., & Krippendorff, K. (2007). Answering the call for a standard reliability measure for coding data. Communication Methods and Measures, 1(1), 77–89. Heckman, C. J., Handorf, E., Darlow, S. D., Yaroch, A. L., & Raivitch, S. (2017). Refinement of measures to assess psychosocial constructs associated with skin cancer risk and protective behaviors of young adults. Journal of Behavioral Medicine, 40(4), 574–582. https://doi.org/10.1007/s10865-017-9825-3 Heron, M., & Anderson, R. N. (2015). Changes in the leading cause of death: Recent patterns in heart disease and cancer mortality. NCHS Data Brief, 254, 1. Hildon, Z., Allwood, D., & Black, N. (2012). Impact of format and content of visual display of data on comprehension, choice and preference: A systematic review. International Journal for Quality in Health Care, 24(1), 55–64. https://doi.org/10.1093/intqhc/mzr072 Hochbaum, G., Rosenstock, I., & Kegels, S. (1952). Health belief model.Washington, DC: United States Public Health Service. Hollands, G. J., Hankins, M., & Marteau, T. M. (2010). Visual feedback of individuals’ medical imaging results for changing health behaviour. Cochrane Database of Systematic Reviews, (1). https://doi.org/10.1002/14651858.CD007434.pub2 144 Hornung, R. L., & Strecher, V. J. (2012). Ultraviolet photography as a skin cancer risk assessment and intervention tool. Journal of the American Academy of Dermatology, 67(4), 785–786. https://doi.org/10.1016/j.jaad.2012.02.016 Houts, P. S., Doak, C. C., Doak, L. G., & Loscalzo, M. J. (2006). The role of pictures in improving health communication: A review of research on attention, comprehension, recall, and adherence. Patient Education and Counseling, 61(2), 173–190. https://doi.org/10.1016/j.pec.2005.05.004 Hovland, C. I., Janis, I. L., & Kelley, H. H. (1953). Communication and persuasion. New Haven, CT: Yale University Press. iMotions. (2016, February 19). Retrieved January 15, 2018, from Facial expression analysis: The complete pocket guide website: https://imotions.com/blog/facialexpression-analysis/ Jackson, S., & Jacobs, S. (1983). Generalizing about messages: Suggestions for design and analysis of experiments. Human Communication Research, 9(2), 169–191. https://doi.org/10.1111/j.1468-2958.1983.tb00691.x Jain, P., Hoffman, E., Beam, M., & Xu, S. S. (2017). Effect of message format and content on attitude accessibility regarding sexually transmitted infections. Health Communication, 32(11), 1376–1384. Jensen, J. D. (2008). Scientific uncertainty in news coverage of cancer research: Effects of hedging on scientists’ and journalists’ credibility. Human Communication Research, 34(3), 347–369. https://doi.org/10.1111/j.1468-2958.2008.00324.x Jensen, J. D. (2011). Addressing health literacy in the design of health messages. In H. Cho (Ed.), Health communication message design: Theory and practice (pp. 171– 190). Thousand Oaks, CA: Sage. Jensen, J. D., Carcioppolo, N., King, A. J., Bernat, J. K., Davis, L., Yale, R., & Smith, J. (2011). Including limitations in news coverage of cancer research: Effects of news hedging on fatalism, medical skepticism, patient trust, and backlash. Journal of Health Communication, 16(5), 486–503. https://doi.org/10.1080/10810730.2010.546491 Jensen, J. D., Carcioppolo, N., King, A. J., Scherr, C. L., Jones, C. L., & Niederdeppe, J. (2014). The cancer information overload (CIO) scale: Establishing predictive and discriminant validity. Patient Education and Counseling, 94(1), 90–96. https://doi.org/10.1016/j.pec.2013.09.016 Jensen, J. D., & Hurley, R. J. (2010). Conflicting stories about public scientific controversies: Effects of news convergence and divergence on scientists’ credibility. Public Understanding of Science, 21(6), 689–704. https://doi.org/10.1177/0963662510387759 145 Jensen, J. D., King, A. J., Carcioppolo, N., & Davis, L. (2012). Why are tailored messages more effective? A multiple mediation analysis of a breast cancer screening intervention. Journal of Communication, 62(5), 851–868. https://doi.org/10.1111/j.1460-2466.2012.01668.x Jensen, J. D., King, A. J., Carcioppolo, N., Krakow, M., Samadder, N. J., & Morgan, S. (2014). Comparing tailored and narrative worksite interventions at increasing colonoscopy adherence in adults 50–75: A randomized controlled trial. Social Science & Medicine, 104, 31–40. https://doi.org/10.1016/j.socscimed.2013.12.003 Jensen, J. D., Krakow, M., Christy, K. R., Ratcliff, C. L., Pokharel, M., & Yale, R. N. (2017, May). Message impact theory: Testing a process-oriented framework for communication evaluation and research. Presented at the 67th annual meeting of the International Communication Association, San Diego, CA. Jones, C. J., Smith, H., & Llewellyn, C. (2014). Evaluating the effectiveness of health belief model interventions in improving adherence: A systematic review. Health Psychology Review, 8(3), 253–269. https://doi.org/10.1080/17437199.2013.802623 Kessels, L. T. E., Ruiter, R. A. C., Brug, J., & Jansma, B. M. (2011). The effects of tailored and threatening nutrition information on message attention. Evidence from an event-related potential study. Appetite, 56(1), 32–38. https://doi.org/10.1016/j.appet.2010.11.139 King, A. J. (2015a). Visual messaging and risk communication. In H. Cho, K. A. Reimer, & K. A. McComas (Eds.), Sage handbook of risk communication (pp. 193–205). Thousand Oaks, CA: Sage. King, A. J. (2015b). A content analysis of visual cancer information: Prevalence and use of photographs and illustrations in printed health materials. Health Communication, 30(7), 722–731. https://doi.org/10.1080/10410236.2013.878778 King, A. J, Carcioppolo, N., Grossman, D., John, K. K., & Jensen, J. D. (2015). A randomised test of printed educational materials about melanoma detection: Varying skin self-examination technique and visual image dose. Health Education Journal, 74(6), 732–742. https://doi.org/10.1177/0017896914558645 Kline, R. B. (2015). The mediation myth. Basic and Applied Social Psychology, 37(4), 202–213. https://doi.org/10.1080/01973533.2015.1049349 Kreuter, M., Farrell, D., Olevitch, L., & Brennan, L. (2000). Tailoring Health Messages. New York: Routledge. Kreuter, M. W., Strecher, V. J., & Glassman, B. (1999). One size does not fit all: The case for tailoring print materials. Annals of Behavioral Medicine, 21(4), 276. https://doi.org/10.1007/BF02895958 146 Kreuter, M. W., & Wray, R. J. (2003). Tailored and targeted health communication: Strategies for enhancing information relevance. American Journal of Health Behavior, 27, S227–S232. Krieger, J. L., & Sarge, M. A. (2013). A serial mediation model of message framing on intentions to receive the human papillomavirus (HPV) vaccine: Revisiting the role of threat and efficacy perceptions. Health Communication, 28(1), 5–19. https://doi.org/10.1080/10410236.2012.734914 Krippendorff, K. (2004). Reliability in content analysis. Human Communication Research, 30(3), 411–433. https://doi.org/10.1111/j.1468-2958.2004.tb00738.x Kutner, M., Greenburg, E., Jin, Y., & Paulsen, C. (2006). The health literacy of America’s adults: Results from the 2003 national assessment of adult literacy. NCES 2006-483. Retrieved from http://eric.ed.gov/?id=ED493284 Leckner, S. (2012). Presentation factors affecting reading behaviour in readers of newspaper media: An eye-tracking perspective. Visual Communication, 11(2), 163–184. Retrieved from ufh. Lee, S., & Lang, A. (2009). Discrete emotion and motivation: Relative activation in the appetitive and aversive motivational systems as a function of anger, sadness, fear, and joy during televised information campaigns. Media Psychology, 12(2), 148– 170. Retrieved from ufh. Leshner, G., Bolls, P., & Wise, K. (2011). Motivated processing of fear appeal and disgust images in televised anti-tobacco ads. Journal of Media Psychology, 23(2), 77–89. https://doi.org/10.1027/1864-1105/a000037 Lester, P. M. (2013). Visual Communication: Images with Messages. Andover, UK: Cengage Learning. Leventhal, H. (1970). Findings and theory in the study of fear communications. Advances in Experimental Social Psychology, 5, 119–186. https://doi.org/10.1016/S00652601(08)60091-X Leventhal, H. (1971). Fear appeals and persuasion: The differentiation of a motivational construct. American Journal of Public Health, 61, 1208–1224. Liberman, A., & Chaiken, S. (1992). Defensive processing of personally relevant health messages. Personality and Social Psychology Bulletin, 18(6), 669–679. https://doi.org/10.1177/0146167292186002 Lim, S., & Reeves, B. (2009). Being in the game: Effects of avatar choice and point of view on psychophysiological responses during play. Media Psychology, 12(4), 348–370. https://doi.org/10.1080/15213260903287242 147 Lipkus, I. M. (2007). Numeric, verbal, and visual formats of conveying health risks: Suggested best practices and future recommendations. Medical Decision Making, 27(5), 696–713. https://doi.org/10.1177/0272989X07307271 Lipkus, I. M., & Hollands, J. G. (1999). The visual communication of risk. Journal of the National Cancer Institute, 25, 149–163. Mahler, H. I. (2014). The role of emotions in UV protection intentions and behaviors. Psychology, Health & Medicine, 19(3), 344–354. https://doi.org/10.1080/13548506.2013.802359 Mahler, H. I. (2015). Interventions to promote sun protection behaviors: What do we know about the efficacy of health- and appearance-based messages and the role of cognitions and emotions? Social and Personality Psychology Compass, 9(5), 238–251. https://doi.org/10.1111/spc3.12173 Mahler, H. I. (2018). The relative role of cognitive and emotional reactions in mediating the effects of a social comparison sun protection intervention. Psychology & Health, 33(2), 235–257. https://doi.org/10.1080/08870446.2017.1310860 Mahler, H. I., Kulik, J. A., Gerrard, M., & Gibbons, F. X. (2007). Long-term effects of appearance-based interventions on sun protection behaviors. Health Psychology, 26(3), 350–360. https://doi.org/10.1037/0278-6133.26.3.350 Mahler, H. I., Kulik, J. A., Gerrard, M., & Gibbons, F. X. (2013). Effects of photoaging information and UV photo on sun protection intentions and behaviours: A crossregional comparison. Psychology & Health, 28(9), 1009–1031. https://doi.org/10.1080/08870446.2013.777966 Maloney, E. K., Lapinski, M. K., & Witte, K. (2011). Fear appeals and persuasion: A review and update of the extended parallel process model. Social and Personality Psychology Compass, 5(4), 206–219. https://doi.org/10.1111/j.17519004.2011.00341.x Manne, S., Jacobsen, P. B., Ming, M. E., Winkel, G., Dessureault, S., & Lessin, S. R. (2010). Tailored versus generic interventions for skin cancer risk reduction for family members of melanoma patients. Health Psychology, 29(6), 583–593. https://doi.org/10.1037/a0021387 Maxwell, S. E., & Cole, D. A. (2007). Bias is cross-sectional analyses of longitudinal mediation. Psychological Methods, 12, 23–44. Mays, D., & Zhao, X. (2016). The influence of framed messages and self-affirmation on indoor tanning behavioral intentions among 18 to 30 year old women. Health Psychology, 35(2), 123–130. https://doi.org/10.1037/hea0000253 McCambridge, J., Kypri, K., & Elbourne, D. (2014). In randomization we trust? There are overlooked problems in experimenting with people in behavioral intervention 148 trials. Journal of Clinical Epidemiology, 67(3), 247–253. https://doi.org/10.1016/j.jclinepi.2013.09.004 McClure, K. J., Puhl, R. M., & Heuer, C. A. (2011). Obesity in the news: Do photographic images of obese persons influence antifat attitudes? Journal of Health Communication, 16(4), 359–371. https://doi.org/10.1080/10810730.2010.535108 McWhirter, J., & Hoffman-Goetz, L. (2014). A systematic review of visual image theory, assessment, and use in skin cancer and tanning research. Journal of Health Communication, 19(6), 738–757. Meczkowski, E. J., Dillard, J. P., & Shen, L. (2016). Threat appeals and persuasion: Seeking and finding the elusive curvilinear effect. Communication Monographs, 83(3), 373–395. https://doi.org/10.1080/03637751.2016.1158412 Messaris, P. (2003). Visual communication: Theory and research. Journal of Communication, 53(3), 551–556. https://doi.org/10.1111/j.14602466.2003.tb02608.x Miller, K. A., Huh, J., Unger, J. B., Richardson, J. L., Allen, M. W., Peng, D. H., & Cockburn, M. G. (2015). Patterns of sun protective behaviors among Hispanic children in a skin cancer prevention intervention. Preventive Medicine: An International Journal Devoted to Practice and Theory, 81, 303–308. Retrieved from psyh. (2015-55783-047) Moser, R. P., Mccaul, K., Peters, E., Nelson, W., & Marcus, S. E. (2007). Associations of perceived risk and worry with cancer health-protective actions. Journal of Health Psychology, 12(1), 53–65. https://doi.org/10.1177/1359105307071735 Nabi, R. L., & Myrick, J. G. (2018). Uplifting fear appeals: Considering the role of hope in fear-based persuasive messages. Health Communication, 1–12. https://doi.org/10.1080/10410236.2017.1422847 National Cancer Institute. (2016, September 8). Retrieved May 26, 2017, from Skin Cancer website: https://www.cancer.gov/types/skin/patient/skin-preventionpdq#section/_16 National Cancer Institute. (2017). Retrieved May 15, 2017, from Cancer Statistics website: https://www.cancer.gov/about-cancer/understanding/statistics Nguyen, M. H., Smets, E., Bol, N., Loos, E. F., & van Weert, J. C. M. (2018). How tailoring the mode of information presentation influences younger and older adults’ satisfaction with health websites. Journal of Health Communication, 23(2), 170–180. Nguyen, M. H., van Weert, J. C. M., Bol, N., Loos, E. F., Tytgat, K. M. A. J., van de Ven, A. W. H., & Smets, E. M. A. (2017). Tailoring the mode of information 149 presentation: Effects on younger and older adults’ attention and recall of online information. Human Communication Research, 43(1), 102–126. https://doi.org/10.1111/hcre.12097 Noar, S. M., Benac, C. N., & Harris, M. S. (2007). Does tailoring matter? Meta-analytic review of tailored print health behavior change interventions. Psychological Bulletin, 133(4), 673–693. https://doi.org/10.1037/0033-2909.133.4.673 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(3), 251–274. https://doi.org/10.1111/j.14682885.2003.tb00292.x O’Keefe, D. J. (2013). The elaboration likelihood model. In J. P Dillard & L. Shen (Eds.), The Sage handbook of persuasion: Developments in theory and practice (2nd ed., pp. 137–149). Thousand Oaks, CA: Sage Publications, Inc. O’Keefe, D. J., & Jensen, J. D. (2006). The advantages of compliance or the disadvantages of noncompliance? A meta-analytic review of the relative persuasive effectiveness of gain-framed and loss-framed messages. Annals of the International Communication Association, 30(1), 1–43. https://doi.org/10.1080/23808985.2006.11679054 Ordoñana, J. R., González-Javier, F., Espín-López, L., & Gómez-Amor, J. (2009). Selfreport and psychophysiological responses to fear appeals. Human Communication Research, 35(2), 195–220. https://doi.org/10.1111/j.1468-2958.2009.01344.x Passyn, K., & Sujan, M. (2006). Self-accountability emotions and fear appeals: Motivating behavior. Journal of Consumer Research, 32(4), 583–589. https://doi.org/10.1086/500488 Pepper, J. K., Cameron, L. D., Reiter, P. L., McRee, A.-L., & Brewer, N. T. (2013). Nonsmoking male adolescents’ reactions to cigarette warnings. PloS One, 8(8), e65533–e65533. https://doi.org/10.1371/journal.pone.0065533 Peters, E. (2012). Beyond comprehension: The role of numeracy in judgements and decisions. Current Directions in Psychological Science, 21(1), 31–35. https://doi.org/10.1177/0963721411429960 Peters, E., Hibbard, J., Slovic, P., & Dieckmann, N. (2007). Numeracy skill and the communication, comprehension, and use of risk-benefit information. Health Affairs, 26(3), 741–748. https://doi.org/10.1377/hlthaff.26.3.741 Peters, G.-J. Y., Ruiter, R. A. C., & Kok, G. (2013). Threatening communication: A critical re-analysis and a revised meta-analytic test of fear appeal theory. Health Psychology Review, 7(sup1), S8–S31. https://doi.org/10.1080/17437199.2012.703527 150 Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. Advances in Experimental Social Psychology, 19, 123–205. https://doi.org/10.1016/S0065-2601(08)60214-2 Politi, M. C., Han, P. K. J., & Col, N. F. (2007). Communicating the uncertainty of harms and benefits of medical interventions. Medical Decision Making, 27(5), 681–695. https://doi.org/10.1177/0272989X07307270 Popova, L. (2012). The extended parallel process model: Illuminating the gaps in research. Health Education & Behavior, 39(4), 455–473. https://doi.org/10.1177/1090198111418108 Potter, R. F., & Choi, J. (2006). The effects of auditory structural complexity on attitudes, attention, arousal, and memory. Media Psychology, 8(4), 395–419. Retrieved from ufh. Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40(3), 879–891. https://doi.org/10.3758/BRM.40.3.879 Quick, B. L., Shen, L., & Dillard, J. P. (2013). Reactance theory and persuasion. In J. P. Dillard & L. Shen (Eds.), The Sage handbook of persuasion: Developments in theory and practice (pp. 167–183). Thousand Oaks, CA: Sage Publications, Inc. Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude Change. The Journal of Psychology, 91(1), 93–114. https://doi.org/10.1080/00223980.1975.9915803 Rosenstock, I. M. (1990). The health belief model: Explaining health behavior through expectancies. In K. Glanz, F. M. Lewis, & B. K. Rimer (Eds.), Health behavior and health education: Theory, research, and practice (pp. 39–62). San Francisco, CA, US: Jossey-Bass. Rothman, A., & Kiviniemi, M. T. (1999). Treating people with information: An analysis and review of approaches to communicating health risk information. Journal of National Cancer Institute Monographs, 25(25), 44–51. Ruiter, R. A. C., Kessels, L. T. E., Peters, G.-J. Y., & Kok, G. (2014). Sixty years of fear appeal research: Current state of the evidence. International Journal of Psychology, 49(2), 63–70. https://doi.org/10.1002/ijop.12042 Schapira, M. M., Nattinger, A. B., & McAuliffe, T. L. (2006). The influence of graphic format on breast cancer risk communication. Journal of Health Communication, 11(6), 569–582. https://doi.org/10.1080/10810730600829916 Schmidt, F. L., & Hunter, J. E. (2014). Methods of meta-analysis: Correcting error and bias in research findings. Thousand Oaks, CA: SAGE Publications. 151 Scholes, D., McBride, C. M., Grothaus, L., Civic, D., Ichikawa, L. E., Fish, L. J., & Yarnall, K. S. (2003). A tailored minimal self-help intervention to promote condom use in young women: Results from a randomized trial. AIDS, 17(10). Retrieved from https://journals.lww.com/aidsonline/Fulltext/2003/07040/A_tailored_minimal_sel f_help_intervention_to.16.aspx Severtson, D. J., & Myers, J. D. (2013). The influence of uncertain map features on risk beliefs and perceived ambiguity for maps of modeled cancer risk from air pollution. Risk Analysis, 33(5), 818–837. https://doi.org/10.1111/j.15396924.2012.01893.x Shen, L., & Coles, V. B. (2015). Fear and psychological reactance. Zeitschrift Für Psychologie, 223(4), 225–235. https://doi.org/10.1027/2151-2604/a000224 Shi, J. (Jolie), & Smith, S. W. (2016). The effects of fear appeal message repetition on perceived threat, perceived efficacy, and behavioral intention in the extended parallel process model. Health Communication, 31(3), 275–286. Retrieved from ufh. Shipp, A. J., & Aeon, B. (2018). Temporal focus: Thinking about the past, present, and future. Current Opinion in Psychology, 26, 37–43. https://doi.org/10.1016/j.copsyc.2018.04.005 Siegel, R. L., Miller, K. D., & Jemal, A. (2017). Cancer statistics, 2017. CA: A Cancer Journal for Clinicians, 67(1), 7–30. https://doi.org/10.3322/caac.21387 Skinner, C. S., Tiro, J., & Champion, V. L. (2015). The health belief model. In K. Glanz, B. K. Rimer, & K. Viswanath (Eds.), Health behavior: Theory, research, and practice (5th ed., Vol. 2, pp. 75–94). Hoboken, NJ: Wiley. Slater, M. D. (2002). Entertainment-education and elaboration likelihood: Understanding the processing of narrative persuasion. Communication Theory, 12(2), 173. Retrieved from ufh. Smerecnik, C. M. R., Mesters, I., Kessels, L. T. E., Ruiter, R. A. C., De Vries, N. K., & De Vries, H. (2010). Understanding the positive effects of graphical risk information on comprehension: Measuring attention directed to written, tabular, and graphical risk information. Risk Analysis, 30(9), 1387–1398. https://doi.org/10.1111/j.1539-6924.2010.01435.x So, J. (2013). A further extension of the extended parallel process model (E-EPPM): Implications of cognitive appraisal theory of emotion and dispositional coping style. Health Communication, 28(1), 72–83. Retrieved from ufh. Solomon, R. L. (1949). An extension of control group design. Psychological Bulletin, 46(2), 137–150. https://doi.org/10.1037/h0062958 152 Sparks, J. V., & Lang, A. (2010). An initial examination of the post-auricular reflex as a physiological indicator of appetitive activation during television viewing. Communication Methods & Measures, 4(4), 311–330. Retrieved from ufh. Spiegelhalter, D., Pearson, M., & Short, I. (2011). Visualizing uncertainty about the future. Science, 333(6048), 1393–1400. https://doi.org/10.1126/science.1191181 Stock, M. L., Gerrard, M., Gibbons, F. X., Dykstra, J. L., Weng, C.-Y., Mahler, H. I., … Kulik, J. A. (2010). Sun protection intervention for highway workers: Long-term efficacy of UV photography and skin cancer information on men’s protective cognitions and behavior: Erratum. Annals of Behavioral Medicine, 39(1), 100– 100. https://doi.org/10.1007/s12160-010-9179-3 Tannenbaum, M. B., Hepler, J., Zimmerman, R. S., Saul, L., Jacobs, S., Wilson, K., & Albarracín, D. (2015). Appealing to fear: A meta-analysis of fear appeal effectiveness and theories. Psychological Bulletin, 141(6), 1178–1204. https://doi.org/10.1037/a0039729 Tate, C. U. (2015). On the overuse and misuse of mediation analysis: It may be a matter of timing. Basic and Applied Social Psychology, 37(4), 235–246. https://doi.org/10.1080/01973533.2015.1062380 Taylor, M. F., Westbrook, D., & Chang, P. (2016). Using UV photoaged photography to better understand Western Australian teenagers’ attitudes towards adopting sunprotective behaviors. International Journal of Adolescent Medicine and Health, 28(1), 45–53. https://doi.org/10.1515/ijamh-2014-0071 Thomas, J., Rompay, V., De vries, P. W., & Van Venrooji, X. G. (2010). More than words: on the importance of picture–text congruence in the online environment. Journal of Interactive Marketing, 24(1), 22–30. Thrasher, J. F., Swayampakala, K., Borland, R., Nagelhout, G., Yong, H.-H., Hammond, D., … Hardin, J. (2016). Influences of self-efficacy, response efficacy, and reactance on responses to cigarette health warnings: A longitudinal study of adult smokers in Australia and Canada. Health Communication, 31(12), 1517–1526. https://doi.org/10.1080/10410236.2015.1089456 Turrisi, R., Hillhouse, J., & Gebert, C. (1998). Examination of cognitive variables relevant to sunbathing. Journal of Behavioral Medicine, 21(3), 299–313. https://doi.org/10.1023/A:1018773101658 Visschers, V. H. M., Meertens, R. M., Passchier, W. W. F., & Vries, N. N. K. de. (2009). Probability information in risk communication: A review of the research literature. Risk Analysis, 29(2), 267–287. https://doi.org/10.1111/j.15396924.2008.01137.x Vraga, E., Bode, L., & Troller-Renfree, S. (2016). Beyond self-reports: Using eye tracking to measure topic and style differences in attention to social media 153 content. Communication Methods and Measures, 10(2–3), 149–164. https://doi.org/10.1080/19312458.2016.1150443 Walsh, L. A., & Stock, M. L. (2012). UV photography, masculinity, and college men’s sun protection cognitions. Journal of Behavioral Medicine, 35(4), 431–442. https://doi.org/10.1007/s10865-011-9372-2 Walsh, L. A., Stock, M. L., Peterson, L. M., & Gerrard, M. (2014). Women’s sun protection cognitions in response to UV photography: The role of age, cognition, and affect. Journal of Behavioral Medicine, 37(3), 553–563. https://doi.org/10.1007/s10865-013-9512-y Waters, E. A., Weinstein, N. D., Colditz, G. A., & Emmons, K. M. (2007). Reducing aversion to side effects in preventive medical treatment decisions. Journal of Experimental Psychology: Applied, 13(1), 11–21. https://doi.org/10.1037/1076898X.13.1.11 Weidman, A. C., Steckler, C. M., & Tracy, J. L. (2017). The jingle and jangle of emotion assessment: Imprecise measurement, casual scale usage, and conceptual fuzziness in emotion research. Emotion, 17(2), 267–295. https://doi.org/10.1037/emo0000226 Weinstein, N. D., Kwitel, A., McCaul, K. D., Magnan, R. E., Gerrard, M., & Gibbons, F. X. (2007). Risk perceptions: Assessment and relationship to influenza vaccination. Health Psychology, 26(2), 146–151. https://doi.org/10.1037/02786133.26.2.146 White, V., Webster, B., & Wakefield, M. (2008). Do graphic health warning labels have an impact on adolescents’ smoking-related beliefs and behaviours? Addiction, 103(9), 1562–1571. https://doi.org/10.1111/j.1360-0443.2008.02294.x Wilkowski, B. M., & Robinson, M. D. (2010). The anatomy of anger: An integrative cognitive model of trait anger and reactive aggression. Journal of Personality, 78(1), 9–38. https://doi.org/10.1111/j.1467-6494.2009.00607.x Williams, A. L., Grogan, S., Clark-Carter, D., & Buckley, E. (2013). Appearance‐based interventions to reduce ultraviolet exposure and/or increase sun protection intentions and behaviours: A systematic review and meta-analyses. British Journal of Health Psychology, 18(1), 182–217. https://doi.org/10.1111/j.20448287.2012.02089.x Witte, K. (1992). Putting the fear back into fear appeals: The extended parallel process model. Communication Monographs, 59(4), 329–349. https://doi.org/10.1080/03637759209376276 Witte, K. (1994). Fear control and danger control: A test of the extended parallel process model (EPPM). Communication Monographs, 61(2), 113–134. https://doi.org/10.1080/03637759409376328 154 Witte, K. (2013). Introduction: Pathways. Health Communication, 28(1), 3–4. https://doi.org/10.1080/10410236.2013.743783 Witte, K., & Allen, M. (2000). A meta-analysis of fear appeals: Implications for effective public health campaigns. Health Education & Behavior, 27(5), 591–615. https://doi.org/10.1177/109019810002700506 Witte, K., Cameron, K. A., Mckeon, J. K., & Berkowitz, J. M. (1996). Predicting risk behaviors: Development and validation of a diagnostic scale. Journal of Health Communication, 1(4), 317–342. https://doi.org/10.1080/108107396127988 Wu, Y. P., Aspinwall, L. G., Conn, B. M., Stump, T., Grahmann, B., & Leachman, S. A. (2016). A systematic review of interventions to improve adherence to melanoma preventive behaviors for individuals at elevated risk. Preventive Medicine, 88, 153–167. https://doi.org/10.1016/j.ypmed.2016.04.010 Yzer, M. (2011). The integrative model of behavioral prediction as a tool for designing health messages. In H. Cho (Ed.), Health communication message design: Theory and practice (pp. 21–40). Thousand Oaks, CA: Sage. Zhang, J. (Mandy), Chen, G. M., Chock, T. M., Wang, Y., Ni, L., & Schweisberger, V. (2016). A psychophysiological study of processing HIV/AIDS public service announcements: The effects of novelty appeals, sexual appeals, narrative versus statistical evidence, and viewer’s sex. Health Communication, 31(7), 853–862. Retrieved from ufh. Zikmund-Fisher, B. J., Ubel, P. A., Smith, D. M., Derry, H. A., McClure, J. B., Stark, A., … Fagerlin, A. (2008). Communicating side effect risks in a tamoxifen prophylaxis decision aid: The debiasing influence of pictographs. Patient Education & Counseling, 73(2), 209–214 Zillmann, D. (1971). Excitation transfer in communication-mediated aggressive behavior. Journal of Experimental Social Psychology, 7(4), 419–434. https://doi.org/10.1016/0022-1031(71)90075-8 Zillmann, D. (1999). Exemplification theory: Judging the whole by some of its parts. Media Psychology, 1(1), 69.