SelfRegulation of Health Behavior Through Selective Exposure to ...

Journal of Communication ISSN 0021-9916

ORIGINAL ARTICLE

To Your Health: Self-Regulation of Health Behavior Through Selective Exposure to Online Health Messages Silvia Knobloch-Westerwick, Benjamin K. Johnson, & Axel Westerwick School of Communication, The Ohio State University, Columbus, OH 43210, USA

Reaching target audiences is of crucial importance for the success of health communication campaigns, but individuals may avoid health messages if they challenge their beliefs or behaviors. A lab study (N = 419) examined effects of messages’ consistency with participants’ behavior and source credibility on selective exposure for 4 health lifestyle topics. Drawing on self-regulation theory and dissonance theory, 3 motivations were examined: self-bolstering, self-motivating, and self-defending. Prior behavior predicted selective exposure across topics, reflecting self-bolstering. Standard-behavior discrepancies also affected selective exposure, consistent with self-motivating rather than self-defending. Selective exposure to highcredibility sources advocating for organic food, fruits and vegetable consumption, exercise, and limiting coffee all fostered accessibility of related standards, whereas messages from low-credibility sources showed no such impact. doi:10.1111/jcom.12055

Reaching target audiences is of crucial importance for the success of health communication campaigns (e.g., Hornik, 2002; Morris, Rooney, Wray, & Kreuter, 2009; Slater, 2004). A lack of exposure is likely the greatest obstacle in these endeavors, making selective exposure pivotal for health communication (e.g., Pease, Brannon, & Pilling, 2006). Various models and approaches aim to explain health information seeking (Case, Andrews, Johnson, & Allard, 2005; Dutta & Bodie, 2008; Griffin, Dunwoody, & Neuwirth, 1999; Johnson & Meischke, 1993). In contrast to prior work, this study uses both self-regulation theory (e.g., Bandura, 1991; Lord, Diefendorff, Schmidt, & Hall, 2010) and cognitive dissonance theory (Festinger, 1957) to derive specific, competing hypotheses on selective health message exposure. While dissonance theory has guided early research into selective use of health messages, the self-regulation perspective presents a new approach to this phenomenon as well as selective exposure in general beyond health communication. To further advance existing knowledge, this

Corresponding author: Silvia Knobloch-Westerwick; e-mail: [email protected] Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

807

Selective Exposure to Health Messages

S. Knobloch-Westerwick et al.

study draws on the selective exposure paradigm and methodology (e.g., KnoblochWesterwick, in press). The research design also incorporates source credibility due to its great importance for persuasive effects (see Pornpitakpan, 2004, for a review and Metzger, Flanagin, & Medders, 2010, for conceptualization) and thus extends selective exposure research in health communication and beyond (Knobloch-Westerwick, in press). Moreover, this study examines consequences of health message exposure, which existing research has hardly ever investigated in connection with selective message use (Knobloch-Westerwick & Sarge, in press). Self-regulation through selective health message exposure

Health communication campaigns aim to improve health behaviors (e.g., Hornik, 2002). As individuals engage in or refrain from healthful behaviors, their selfregulation processes will involve information use pertaining to the related health topics. Such information usually implies behavioral standards (e.g., ‘‘exercise 5 days a week’’), resulting in a positive self-evaluative reaction (Bandura, 1991) if that standard is met or negative self-evaluative reaction if one’s behavior falls short. Hence, selective exposure may be crucial for health behavior: Theories on self-regulation argue that individuals selectively activate and deactivate behavioral standards (e.g., Bandura, 1991; Lord et al., 2010), and an important route of performing such activation and deactivation is attending selectively to information pertaining to standards of healthful behavior. Specifically, messages that promote health behaviors likely help fostering these behaviors; messages that warn against these behaviors likely aid abstaining from these behaviors. On the basis of this self-regulation perspective, selective exposure to health messages may be utilized for both maintenance of existing behaviors (self-bolstering) and for initiation or fostering of behavior change (self-motivating). While much self-regulation theorizing focuses on behavioral change, the maintenance of behaviors is also of utmost relevance (Rothman, Baldwin, Hertel, & Fugelstad, 2011), in particular for health behaviors. Accordingly, both frequent engagement in a healthful behavior and falling short from a health behavioral standard could foster exposure to messages promoting the behavior. The former pattern would aid behavior maintenance and also elicit positive self-evaluative reactions during exposure: The more individuals engage in certain health behaviors, the more time they spend with messages promoting these health behaviors (self-bolstering) (H1). The latter pattern would result from a drive to motivate oneself to change toward more healthful behavior: The more individuals fall short of perceived standards for certain health behaviors, the more time they spend with messages promoting these health behaviors (self-motivating) (H2a). Yet self-regulation does not always succeed in working toward healthier behaviors (e.g., Baumeister & Heatherton, 1996). Depending on their stage of health behavior change (Prochaska, DiClemente, & Norcross, 1992; Weinstein, 1988), individuals may not be willing or ready to change in the first place—because of competing life demands, high perceived costs of changing (e.g., discomfort from quitting 808

Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association



smoking), or perceived lack of efficacy to achieve the targeted standards. Under these circumstances, individuals may respond defensively to health messages (Liberman & Chaiken, 1992) because exposure to them would create negative self-evaluative reactions (Bandura, 1991) if current behavior falls short of behavioral standards. The associated exposure patterns would be avoidance of messages that promote the health behavior. Avoiding dissonance from health message exposure

The first studies of selective exposure to health messages tested the proposition that messages which challenge one’s behaviors (i.e., smoking) are avoided. Historically, selective exposure research focused on such a defensive bias—individuals are thought to prefer messages aligned with pre-existing views and behaviors, per Festinger’s theory of cognitive dissonance (1957). It should be noted that H1 also aligns with Festinger’s theorizing, whereas H2a does not. Political communication research during 1960–1980 yielded inconsistent findings regarding the defensive bias, but the Internet era has revived the interest (e.g., Donsbach & Mothes, 2012). While consensus now exists regarding a politically motivated defensive bias in selective exposure, scholars acknowledge that individuals may be motivated to attend to counterattitudinal, challenging messages if they carry informational utility and aid in adapting to social or other circumstances (e.g., Knobloch-Westerwick & Kleinman, 2012; Turner, Yao, Baker, Goodman, & Materese, 2010). For health communication, a defensive exposure bias has also been investigated. Specifically, Feather (1962, 1963) and Brock (1965) found that no less reading interest was indicated for messages that challenged habitual health behavior. In contrast, by having participants press a button to receive a recorded message more clearly, Brock and Balloun (1967) demonstrated avoidance of behavior-challenging messages. Additionally, in Weinstein’s work (1979), participants could request either threatening or reassuring cancer information; although threatening information was generally preferred over reassuring information, participants also favored information consistent with their own views. Further, Weinstein (1985) invited participants to request an article about coffee cancer risk and found no avoidance among coffee drinkers. Bertrand’s (1979) field study in a waiting room showed that neither smokers nor hypertensives gazed less at short health films on smoking and hypertension. Recently, Hwang (2010) found no defensive bias in antismoking message exposure in survey data. Hence, avoidance of behavior-challenging health messages was only demonstrated by Brock and Balloun (1967). Yet all outlined studies except Weinstein’s used smoking as a topic and are now somewhat dated. The present research tests the assumption that guided this classic research through a hypothesis that competes with H2a by proposing the opposite pattern: The more individuals fall short of perceived standards for health behaviors, the less time they spend with messages promoting these health behaviors (self-defending) (H2b). Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

809



Selective exposure and source credibility

Various explanations have been discussed for a lack of a defensive bias in selective exposure, as in the research reviewed above. One prominent consideration suggested that low argument strength or source credibility allows recipients to easily refute a message (Festinger, 1964; Lowin, 1967). Lowin’s (1967) approach-avoidance model proposed that exposure to attitude-challenging messages occurs when these are easily discredited (Lowin, 1967, 1969). Psychological research (e.g., Kessels, Ruiter, & Jansma, 2010; Liberman & Chaiken, 1992) has shown that individuals engage in biased processing of personally relevant health messages, in which threatening message parts are evaluated more critically while reassuring parts are evaluated less critically. Thus, if personally relevant health messages do attract recipients, those recipients are subsequently likely to engage in defensive processing (e.g., Brock & Balloun, 1967; Hwang, 2010; Kleinhesselink & Edwards, 1975). On basis of this rationale, health messages may be accessed even if they contain threatening or behavior-challenging information, as long as they come from low-credibility sources and are thus not very persuasive. This moderating impact of source credibility is particularly relevant in the Internet age, because much health information exposure now occurs online (Fox, 2011) and the expertise of related online sources varies greatly (e.g., Dutta-Bergman, 2004b; Hu & Sundar, 2010). Thus the source credibility associated with a health message should moderate its persuasiveness such that messages from low-credibility sources are more easily refuted and thus avoidance may be less important: The effect suggested in H2a/b will be stronger for messages associated with high-credibility sources (H3). The selective exposure paradigm

Clarification of the term selective exposure is in order before examining exposure consequences. The present work draws on a paradigm that defines selective exposure as any observable bias of exposure to available communication content, captured through behavioral measures of selective media use (for a review, see Knobloch-Westerwick, in press). The term is thus broader than information seeking, which pertains to goal-directed, intentional search for information (Dutta-Bergman, 2004b)—selective exposure also includes experientially driven exposure motivations. A variety of theoretical notions (e.g., informational utility and mood management; Knobloch-Westerwick, in press) may serve to derive specific predictions for selective exposure—operationalizations vary accordingly and include exposure time or the difference between attitude-consistent and counterattitudinal message choices (e.g., Knobloch-Westerwick & Kleinman, 2012; Turner et al., 2010). Research within the paradigm builds on the following assumptions. First, media users are largely unaware of their message choices motivations. Further, message choices typically occur in a rather casual fashion and, as a result, are not fully accessible through later recall. Accordingly, survey measures relying on self-reported retrospections of earlier media choices and introspections regarding motivations for message selections are highly problematic in light of these assumptions. Moreover, these measures may be biased 810




due to social desirability of certain media choices (e.g., Prior, 2009) and respondents’ drive for a consistent self-image. Hence, selective exposure research differs conceptually and methodologically from research that relies on survey self-reports. The vast majority of research on health information seeking used survey studies (Rains, 2007; Ruppel & Rains, 2012; Weaver et al., 2010; Ye, 2010). In contrast, research within the paradigm draws on measures of unobtrusive observation of actual media choice behavior. However, a small number of health communication studies included unobtrusive, observational measures of selective message use (although some of them did not explicitly focus on selective exposure and were primarily concerned with responses to health threats). These studies yielded that exemplification, efficacy, and extent of a threat influence selective exposure (Hastall & Knobloch-Westerwick, 2013; Knobloch-Westerwick & Sarge, in press; Rimal & Real, 2003; Turner, Rimal, Morrison, & Kim, 2006). But generally, health communication research largely neglects measurement of actual exposure (Morris et al., 2009). The consequences of health message exposure are, of course, at the heart of health communication research and ideally should be examined through rigorous, observational measures of exposure behavior. Consequences of selective exposure to health messages

The considerations above regarding self-regulation build on the notion that individuals facilitate selective activation of behavioral standards through selective exposure to related information. Accordingly, selective exposure should not only reflect the associated motives, but also influence the activation and thus accessibility of these standards, which is inversely related to the time with which individuals retrieve the behavioral standard from memory. Further, the faster individuals respond when reporting an attitude, the more strongly they hold the attitude and the greater is the predictive value of the attitude for actual behavior (e.g., Fazio, 2001). Behavioral standards can be seen as attitudes toward behaviors (Fazio, 2007): For example, the attitude ‘‘eating organic food is healthy’’ entails an evaluation of an object (eating organic food) and connects to a behavioral standard to purchase and consume (more) organic food. The associated accessibility has significant predictive value for standard pursuit (Bargh, Gollwitzer, Lee-Chai, Barndollar, & Trötschel, 2001; Shah & Kruglanski, 2003), that is, actual organic food consumption. Thus, behavioral standard accessibility is thought to be a key mediator for effects of health message exposure on behavior. Given this crucial role, the following hypothesis will be tested: Selective exposure to messages pertaining to health behaviors fosters the accessibility of standards regarding these behaviors (H4). These impact patterns should again depend on source credibility: The impact suggested in H4 is stronger for messages from high-credibility sources compared to messages from low-credibility sources (H5). Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

811



Method Overview

A lab study (N = 419) served to test the hypotheses on selective exposure to health messages. Messages were presented as online search results regarding four health topics: organic food, coffee, fruit and vegetable consumption, and physical exercise. Stimuli were manipulated in a 4 × 2 × 2 all within-subjects design, with health topics as a four-step factor and source credibility (low vs. high) and issue stance (promoting vs. opposing health behavior) as two-step factors. For each topic, four search results were displayed with a brief preview on a search results overview; full texts associated with the search results were accessible via hyperlink. Two of these messages promoted consumption or relevant behavior, while two suggested avoiding it. Further, within each pair, one message was associated with a low-credibility source and one with a high-credibility source. Topics, sources, and search result previews were selected based on pretests (see Appendices A–D for details). The browsing time span was 2 minutes for each topic, during which participants perused the overview and actual texts while software logged their exposure behavior in seconds. Before and after the selective exposure task, measures on health perceptions and behaviors were collected. Participants

A total of 419 participants provided valid, complete data in exchange for extra course credit. They were recruited online by instructors at a large university in the Midwestern United States. The sample was 53.0% female, 69.5% White, 12.6% Black, 12.4% Asian, 1.7% Hispanic, and 3.8% other. The average age was 21.11 (SD = 3.11). The participants were heavy Internet users—92% said they were online every day and 73% indicated three or more hours of daily Internet use. Procedure Setting and sequence

All sessions were completed in a computer lab, at private workstations. The entire computerized procedure was set up with Microsoft Silverlight. After providing informed consent, participants first completed the baseline section with questions about health topic perceptions, which typically lasted 5 minutes. Second, participants did a distractor task, which took about 15 minutes and consisted of browsing news unrelated to health matters and a questionnaire. Third, participants completed a selective exposure task for 8 minutes. The fourth and final part was a postexposure questionnaire that took about 12 minutes. Participants were then debriefed and awarded extra credit. Baseline section

Participants responded to cues regarding the four target topics, embedded in eight distractor topics, to measure standard accessibility and importance. Further, they completed measures of media use and demographics. 812




Distractor section

Participants then switched to distractor tasks for 15 minutes. During this section, they browsed an online news magazine featuring nonhealth stories, regarding topics like identity theft, nuclear power infrastructure, burglary, and gas prices. After 4 minutes of this task, they evaluated a number of news headlines from the magazine they just saw and then completed an inventory regarding their trait coping styles. Selective exposure section

Participants were informed that they would view online search results ‘‘that a news portal turned up for several health topics’’ and ‘‘browse through the available articles and read whatever you find interesting.’’ During the selective exposure task, participants were presented with results pages for the search term ‘‘organic food,’’ followed by ‘‘coffee and health,’’ ‘‘eating vegetables,’’ and ‘‘working out.’’ Page placement of articles was randomized within each topic’s results page to avoid sequence effects. For each topic, the overview page and the four associated texts were accessible for 2 minutes. This browsing time span was chosen based on an analysis of a large transaction log dataset that showed that 75% of search sessions in a consumer health information portal (as mimicked in our experiment) take 122 seconds or less and consist of 2.16 or fewer queries (Wolfram, Wang, & Zhang, 2009, p. 902). During browsing, participants could examine search results previews (headlines and leads) and click through to read any of the individual articles; they could go back to the search results page to click another article or return to an article as often as they wanted. After 2 minutes had elapsed, a pop-up button appeared stating, ‘‘The time to read the articles is elapsed. Please click OK.’’ The participant then continued to a new results page for the next topic. Postexposure section

After the last browsing period, participants first completed source recall questions. Then standard accessibility measures were readministered. Health behaviors and perceived behavior standards regarding target topics were ascertained last. Stimuli Search results pages

During the selective exposure task, participants viewed a website consisting of search results for health terms (see Figure 1 for an illustration). Sets of search results were specific to each health topic and featured four articles each. Articles were manipulated in a 2 × 2 (credibility × issue stance) within-subjects design, with the four topics displayed sequentially as an additional within-subjects factor. For each of the topics—organic food, coffee, fruits and vegetables, and exercise—two articles promoting consumption or the behavior were presented (one with a high-credibility source, one with a low-credibility source), and two articles opposing consumption or the behavior were presented (likewise, with one high-credibility and one lowcredibility source). The search results page was patterned after a popular web portal and news site not typically associated with health information. Finally, a prominent Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

813



Figure 1 Example of search results page.

part of the linked preview for each article was source information placed immediately above the article’s headline (see illustration in Figure 1, specific sources are listed in Appendix B). Sources

Selected sources (see Appendices A and B) were assigned to articles with a Latin square approach, ensuring that each issue stance featured both a high- and a low-credibility source, and that each source was associated with just a single article for any given participant. Articles

Articles were chosen to reflect either support or opposition to the health behavior in question. Using claims and arguments from real online articles, the stimuli argued for or against organic food, coffee, fruits and vegetables, and exercise (see Appendix C for headlines and leads). All articles were adapted from real news and advocacy articles. They were edited for similar style, structure, and length (headline and lead word count M = 29.13, SD = 0.34, article body word count, M = 714.75, SD = 25.22). Measures Selective exposure

The software unobtrusively logged any link that was clicked, so that selective exposure to web pages was recorded in seconds. Variables for selective exposure were generated such that, for each of the four topics, (a) exposure time to a message promoting a 814




behavior from a high-credibility source, (b) exposure time to a message promoting a behavior from a low-credibility source, (c) exposure time to a message suggesting avoidance from a high-credibility source, and (d) exposure time to a message suggesting avoidance from a low-credibility source were captured. Selective exposure to all promotion messages was the sum of (a) and (b), while selective exposure to messages opposing a health behavior was the sum of (c) and (d), respectively. The same logic applied to selective exposure indicated by message choice—that is, whether an article was clicked on at all. For example, for coffee, exposure to the message beginning with the heading and lead ‘‘Guilt-Free Pleasure of Coffee—With coffee shops seemingly on every corner, and a continued increase in American coffee consumption, the news about coffee’s effects on health is surprisingly good’’ was categorized as exposure to a message promoting a behavior, whereas exposure to the message beginning with ‘‘Coffee Is Addictive, Mind-Altering: Caffeine addicts may try their best to give up their coffee habit—but usually are not able to, even when it threatens their very well-being’’ was categorized as exposure to a message suggesting avoidance of that behavior. Categorization of exposure to messages from low- versus high-credibility sources was simply defined based on association with displayed sources representing low versus high credibility, as established in the pretest (see Appendices A and B). Interestingly, preliminary analyses showed that information from high-credibility sources was preferred overall.1 Standard accessibility

A response time task was administered to capture the accessibility (Fazio, 1995) of health behavioral standards—the lower participants’ response reaction time, the greater is their accessibility. The application measured response times in milliseconds. A series of instructions first explained the procedure to participants. A practice session was used to familiarize participants with pressing keys to ‘‘classify items as quickly as you can while making as few mistakes as possible.’’ The ‘‘Z’’ key (left) was always used to denote categorization as ‘‘bad for your health,’’ and the ‘‘/’’ key (right) was always used to denote categorization as ‘‘good for your health.’’ Participants were asked to place the index fingers above these keys, and to focus on accuracy and speed. The trial session utilized six clearly valenced adjectives (e.g., ‘‘Marvelous,’’ ‘‘Painful’’) in a randomized sequence. The target word to be categorized appeared in large black text in the upper center of the screen. ‘‘Bad’’ and ‘‘Good’’ appeared in large dark red text in the bottom half of the page, offset to the left and right, respectively. After this trial session, participants were instructed that, ‘‘In the following, you will be asked about topics where people can have very different opinions. Please keep in mind that there are no ‘right’ or ‘wrong’ answers for these questions and that we are only interested in your personal views and various aspects of your opinion.’’ They were then informed that they would be categorizing health topics, where ‘‘Bad’’ meant ‘‘bad for your health’’ and ‘‘Good’’ meant ‘‘good for your health.’’ The response time task began with a series of eight distractor health topics (e.g., ‘‘Red meat,’’ Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

815



‘‘Artificial sweeteners’’). These were followed by the topics of interest, ‘‘Organic food,’’ ‘‘Coffee,’’ ‘‘Fruits and vegetables,’’ and ‘‘Exercise,’’ which were presented in randomized sequence. Related descriptive statistics are reported in Appendix E. Importance

Seven-point scales were used to measure health topic importance from not at all important to extremely important (see descriptive statistics in Appendix E). Health behaviors

Additionally, participants reported their own health behaviors. For example, they indicated how much of the food they eat in a typical week is organic, ranging from 0 to 100% in 10-percent increments. In addition to the target topics, the distractor topics were included as with the previous measures. For coffee, vegetables, and exercise, the 10-point scale referred to servings per day or frequency per week (see descriptive statistics in Appendix E). Perceived standards

The same 10-point scales used to report behavior served to gauge perceived expert recommendations. Participants indicated how frequently medical experts would recommend each behavior or type of consumption (see descriptive statistics in Appendix E). Standard-behavior discrepancies

To operationalize the discrepancy between perceived standards in recommended health behaviors and actual behavior, the difference between perceived expert recommendation and actual behavior was computed for each of the four target topics (see descriptive statistics in Appendix E). The correlations between these two variables ranged between .25 and .49 (ps < .001) across the four topics. Results Impacts on selective exposure to health messages

Four regression analyses were conducted to address H1 and H2a/b—one for each of the four health topics. Within each topic, selective exposure to messages promoting behaviors served as dependent measure, while the related health behavior frequency and standard-behavior discrepancy were used as predictor variables. The results are reported in the first section of Table 1. The impact pattern was very uniform across topics, supporting H1 and H2a. The more individuals engaged in health behaviors, the more time they spent with messages promoting these health behaviors, which aligns with the notion of self-bolstering. Further, the more individuals fell short of perceived standards in health behaviors, the more time they allotted to content promoting that behavior, which suggests they were seeking to motivate themselves to engage more in those behaviors. By the same token, no evidence for H2b emerged 816




Table 1 Impacts of Health Behavior Frequency (Self-Bolstering) and Standard-Behavior Discrepancy (Self-Motivating) on Selective Exposure to Health Messages Health Behavior Frequency Messages promoting behavior (s) Organic food .21*** Coffee .26*** Fruit and vegetables .12* Exercise .27*** Difference promoting/opposing messages(s) Organic food .19** Coffee .28*** Fruit and vegetables .11* Exercise .24*** Difference promoting/opposing messages (ct.) Organic food .21*** Coffee .30*** Fruit and vegetables .09+ Exercise .27*** Messages opposing behavior (s) Organic food −.15** Coffee −.28*** Fruit and vegetables −.10+ Exercise −.21***

Standard-Behavior Discrepancy

Adjusted R2

.19*** .39*** .17** .34***

.044*** .097*** .018** .045***

.17** .40*** .15** .33***

.032*** .100*** .014* .042***

.10* .40*** .13* .37***

.036*** .107*** .009+ .055***

−.13* −.39*** −.13** −.31***

.019** .095*** .010* .037***

Note: Coefficients with three asterisks are significant at p < .001, with two at p < .01, with one at p < .05, with plus sign at p < .10. The first two and last sections report dependent variables measured in seconds, the third section uses dependent variables measured as the difference between the number of promoting articles selected and number of opposing articles selected.

for defensive avoidance of messages that encouraged a health behavior as a result of not meeting perceived related standards. These result patterns remained the same in analyses that utilized different operationalizations of selective exposure (e.g., Turner et al., 2010)—the difference in exposure times allotted to messages promoting versus opposing the behaviors as well as the difference of number of selected messages promoting versus opposing the behaviors, as reported in the second and third sections of Table 1. Next, these regression models were extended to address two alternative explanations for these findings, which would not align with H1 and H2. First, the same regression models were run but using selective exposure to messages arguing against the health behaviors as dependent variables. These analyses served to address the following competing explanation for the findings: People with higher levels of standard-behavior discrepancy were more likely to be exposed not only to messages promoting the behavior but also to messages suggesting avoidance, with the latter Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

817



exposure type possibly balancing out the former and thus negating dissonance. As the findings in the lowest section of Table 1 show, the opposite emerged. Hence, this alternative explanation was ruled out, with these findings further contradicting H2b. Second, ratings for importance for the respective topic were added to all regression analyses reported above. These extended analyses served to examine the following competing explanation: People who engage in a certain health behavior more, because they consider the health topic as particularly important and are more involved with the topic, may have spent more time with health promotion messages due to this greater involvement (e.g., per complementarity theory; Dutta-Bergman, 2004a; Tian & Robinson, 2008) and not because they are engaging in self-bolstering. The results of the regression analyses remained virtually the same as the beta for importance as predictor never approached significance. Thus this competing explanation was ruled out as well. Addressing H3 specifically, the initial regression models were applied to exposure to messages with low versus high credibility. The patterns, however, remained largely the same as for the first four regression models, suggesting that source credibility did not affect impacts of health behavior and standard-behavior discrepancy on selective exposure. Thus H3 was not supported. Selective exposure impacts on accessibility

Hypothesis 4 was tested through four regression analyses, one for each health topic. The predictors were selective exposure to all messages promoting each behavior (measures of exposure to messages on avoiding the behavior were not included to avoid multicollinearity). The dependent variable was accessibility of standards after exposure. Within each topic, the analyses controlled for baseline accessibility and importance. These analyses yielded significant impacts of exposure to messages promoting organic food consumption (β = .17, p < .001; adjusted R2 = .061, p < .001) and of exposure to messages promoting exercise (β = .11, p = .019; adjusted R2 = .062, p < .001) on accessibility of the related standard. These results provide some support for H4, but the impact is not consistent across topics. Similarly, H5 was addressed through regression analyses, one for each topic, but differentiating by source credibility. These regression analyses examined effects of selective exposure to (a) messages promoting the behavior from high-credibility sources, (b) messages promoting the behavior from low-credibility sources, and (c) messages suggesting avoiding the behavior from high-credibility sources on accessibility of standards after exposure. Again, exposure to messages on behavioral avoidance from low-credibility sources was not included to avoid multicollinearity. Within each topic, the analyses controlled for baseline accessibility and importance. Table 2 summarizes the findings. Selective exposure to messages promoting behaviors from high-credibility sources uniformly increased accessibility, except that for coffee consumption the message on behavior avoidance from a high-credibility source (on the common notion that coffee consumption should be curtailed) fostered accessibility. These findings corroborate H4 in conjunction with H5, as 818




Table 2 Selective Exposure Impacts on Behavioral Standard Accessibility by Message Type (Promoting Behavior or Suggesting Avoidance) and Source Credibility (Beta Weights) Selective Exposure Measures

Topic Organic food Coffee Fruit and vegetables Exercise

Control Variables

Message Promoting Behavior, HighCredibility Source

Message Promoting Behavior, LowCredibility Source

Message Suggesting Avoidance, HighCredibility Source

Baseline Accessibility

Attitude Importance

Adjusted R2

.15** — .12*

— — —

— .12* —

−.10* −.35*** −.21***

.17*** — —

.075*** .138*** .073***

.12*

—

—

−.16**

—

.041*

Note: Beta weights with one asterisk are significant at p < .05, with two asterisks at p < .01, with three asterisks at p < .001. A dash indicates that coefficients were not significant. Measures for selective exposure to avoidance-suggesting messages from low-credibility sources were not included in the model to avoid multicollinearity.

the impacts of selective exposure emerged only for messages from high-credibility sources. Discussion

The present investigation examined how online users attend selectively to health information on the Internet through unobtrusive observation of exposure. In the context of an online search, participants could sample from messages that either promoted certain health behaviors or suggested avoiding these behaviors. Further, the messages were associated with either low- or high-credibility sources. Preliminary analyses showed that information from high-credibility sources was preferred overall. The first set of hypotheses pertained to factors affecting selective exposure. In line with H1, more frequent engagement in the portrayed behaviors fostered selective exposure to messages that promoted these behaviors, reflecting self-bolstering. Further, H2a on self-motivating through exposure was supported: The greater the discrepancy between behavior frequency and a perceived behavioral standard, the more time was spent on these messages. This pattern suggests that participants not meeting perceived recommendations aimed to motivate themselves to consume more organic food, or to consume more fruits and vegetables, or to drink more coffee, or to exercise more often—if they felt they fell short of the recommended frequency of these behaviors. At the same time, if participants felt they drank too much coffee, their discrepancy was negative and they would, on average, spend less time with messages promoting coffee consumption. Given that H2a was supported, the competing hypothesis H2b was not corroborated, as no evidence of self-defensive avoidance of behavior-advocating messages emerged. Further, no evidence emerged that the credibility of the sources associated with the available messages moderated these patterns, which leaves H3 unsupported. Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

819



While the concept of self-bolstering for behavior maintenance was derived from self-regulation theory, the related finding for H1 also aligns with cognitive dissonance theory: The more often individuals engage in a health behavior, the more time they spent with messages promoting that behavior and the less time they spent with messages advocating against it, as the latter likely evokes dissonance. It is difficult to disentangle whether this exposure pattern resulted from efforts to foster that habit (self-regulation view) or to avoid discomfort from behavior-challenging messages (cognitive dissonance view). It is possible that (a) situational dissonance from exposure to a behavior-challenging message has different implications than (b) pre-existing dissonance from a standard-behavior discrepancy. Possibly, individuals are motivated to avoid the first type of dissonance, which adds to impacts on selective exposure stemming from the behavior maintenance motive of self-bolstering, but may seek out messages to increase the second type of dissonance (if topic importance and dissonance are high) so that it can get resolved through behavior change (see Festinger, 1957, pp. 127–129). It is desirable that future research will extricate the roles of behavior maintenance versus dissonance avoidance. It appears, however, that the theoretical frameworks show an interesting overlap in this regard: Individuals avoid messages that challenge behaviors they wish to maintain because they do not wish to deactivate the related behavioral standard through cognitive dissonance pertaining to the behavior. Regarding hypotheses on selective exposure consequences, only partial support emerged for H4, which suggested that selective exposure increases the accessibility of health behavioral standards. Yet when considering source credibility in line with H5, selective exposure consistently enhanced the accessibility of predominant standard perceptions, because only exposure to messages from high-credibility sources fostered accessibility of standards for such benefits of organic food, reduced coffee consumption, and fruits and vegetables consumption, as well as physical exercise. Rendering an attitude more salient and thus more accessible leads to greater behavioral impacts of that attitude (Fazio, 1990; Fazio & Roskos-Ewoldsen, 2005). Individuals who perceive a gap between their actual health behavior and what they believe they should be doing may utilize this phenomenon to motivate themselves toward behavior improvement through selective exposure to suitable messages. Interestingly, no support for any defensive avoidance per H2b emerged in this study. Arguably, the present sample of college students, a comparatively healthy population (Grace, 1997) that frequently uses the Internet for health information search (Escoffery et al., 2005), felt little necessity to avoid behavior-challenging messages. These participants likely did not perceive the messages on health lifestyle topics as immediately threatening and thus were more guided by intentions to improve their behaviors. The featured health behaviors were also all relatively easy to change. These considerations may explain why no defensive avoidance was evident in this study and also why source credibility did not affect selective exposure. Interestingly, however, source credibility mattered for selective exposure implications 820




for accessibility of health behavioral standards. Thus, the sources and their credibility did not go unnoticed. Accordingly, messages from high-credibility sources will also have greater impact on actual behavior through increased accessibility (Fazio, 1990). The present investigation has limitations with regard to the nature of the sample and the health topics. Further, the homogeneity of credibility levels among search results previews within a topic could still be improved, although no significant differences emerged in the stimuli pretest. Notwithstanding these limitations, the methodological approach of tracking actual selective exposure to health messages unobtrusively is well suited for gaining more insights into how health information is selected, processed, and eventually applied to behavioral decisions. With the current results, it is delightful to see that behavior-challenging message exposure does happen: Individuals seek out messages that aid them in bolstering good health habits and in changing toward better health habits. The latter finding contrasts with dissonance theory assumptions that guided early health communication research into selective exposure. It will be important to learn more about the circumstances under which individuals are motivated to seek out information to improve their health behaviors. Future research should study how the impacts of behavior frequency and standardbehavior discrepancy on selective health message exposure depend on individuals’ stage of health behavior change (Prochaska et al., 1992; Weinstein, 1988) and efficacy (Bandura, 1977) signaled by a message or perceived by the individual. The motivations of self-bolstering and self-motivating that were reflected in the presented selective exposure patterns may be relatively common in everyday media use, despite lack of prior research about them. Certain media outlets likely even base their business model on recipients’ seeking to bolster and enhance their health behaviors through selective message exposure. For example, consumers of women’s health and fitness magazines probably purchase these magazines because they find them helpful to keep up their weight loss and weight management efforts (KnoblochWesterwick & Crane, 2012) by reading the ample weight loss messages in these outlets (Willis & Knobloch-Westerwick, in press). The activation of health behavior standards from selective exposure could be an effective route for lasting influences. Indeed, empirical research has shown that selective exposure to weight loss messages influences behavior assessed with a 2-week delayed measure (Knobloch-Westerwick & Sarge, in press). As it stands, examining health message impacts from a selfregulation perspective that accounts for selective exposure and accessibility processes is a very promising route for further insights to guide health communication efforts. Acknowledgment

The authors wish to thank Yiqi Chen for her help with data collection administration. Notes 1 Descriptive analyses showed that participants spent an average of 19.06 s (SD = 15.82) on each results page, with some differences in time spent on results page by topic, F(3, 1269) Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

821



= 9.98, p < .001, ηp 2 = .02. On average, participants read 6.90 (SD = 3.21) articles out of the 16 available over the entire selective exposure task. This did not vary significantly across topics, F(3, 1269) = 0.27, p = .85. Among articles selected for reading, participants spent an average of 72.77 s (SD = 30.90) reading. Across all articles, average reading time was M = 26.17 s (SD = 4.30). Also for descriptive purposes, an ANOVA with 16 selective exposure times as repeated measures—differentiated by message type (promoting behavior vs. suggesting avoidance), source credibility (low vs. high), and topic (organic food, coffee, fruits and vegetables, exercise) as three within-subjects factors—was conducted. It showed that exposure to articles overall differed by topic, F(3, 1254) = 8.80, p < .001, ηp 2 = .021. Subsequent analyses showed that organic food articles garnered significantly (p < .05 in paired t-tests) less exposure time than all other topics; coffee and fruits and vegetables induced equivalent exposure lengths, while exposure to exercise articles was significantly longer than for any other topic (see means in Appendix E). The ANOVA further showed that, overall, message type influenced exposure time, F(1, 418) = 98.75, p < .001, ηp 2 = .191, with longer exposure to messages on behavior avoidance (M = 253, SD = 99) than to messages promoting a behavior (M = 164, SD = 94). However, both the impacts of topic and of message type were further qualified, as an interaction between the two occurred, F(3, 1254) = 32.95, p < .001, ηp 2 = .073, which resulted from the fact that the overall preference for messages on behavior avoidance did not emerge for the topic exercise while it was significant for the three other topics in subsequent tests. Thus, message type did not have a uniform impact across topics. Source credibility yielded an impact on exposure, F(1, 418) = 10.69, p = .001, ηp 2 = .025, which applied uniformly across all four topics. Messages from high-credibility sources generally induced longer exposure times (M = 224, SD = 103) than messages from low-credibility sources (M = 193, SD = 101).

References Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84(2), 191–215. doi: 10.1037//0033-295X.84.2.191 Bandura, A. (1991). Social cognitive theory of self-regulation. Organizational Behavior and Human Decision Processes, 50(2), 248–287. doi: 10.1016/0749-5978(91)90022-L Bargh, J. A., Gollwitzer, P. M., Lee-Chai, A., Barndollar, K., & Trötschel, R. (2001). The automated will: Nonconscious activation and pursuit of behavioral goals. Journal of Personality and Social Psychology, 81(6), 1014–1027. doi: 10.1037/0022-3514.81.6. 1014 Baumeister, R. F., & Heatherton, T. F. (1996). Self-regulation failure: An overview. Psychological Inquiry, 7(1), 1–15. doi: 10.1207/s15327965pli0701_1 Bertrand, J. T. (1979). Selective avoidance on health topics: A field test. Communication Research, 6(3), 271–294. doi: 10.1177/009365027900600302 Brock, T. C. (1965). Commitment to exposure as a determinant of information receptivity. Journal of Personality and Social Psychology, 2(1), 10–19. doi: 10.1037/h0022082 Brock, T. C., & Balloun, J. L. (1967). Behavioral receptivity to dissonant information. Journal of Personality and Social Psychology, 6(4), 413–428. doi: 10.1037/h0021225 Case, D. O., Andrews, J. E., Johnson, J. D., & Allard, S. L. (2005). Avoiding versus seeking: The relationship of information seeking to avoidance, blunting, coping, dissonance, and related concepts. Journal of the Medical Library Association, 93(3), 353–362. 822




Donsbach, W., & Mothes, C. (2012). The dissonant self: Contributions from dissonance theory to a new agenda for studying political communication. Communication Yearbook, 36, 3–44. Dutta, M. J., & Bodie, G. D. (2008). Web searching for health: Theoretical foundations and connections to health related outcomes. Web Search: Springer Series in Information Science and Knowledge Management, 14(4), 235–254. doi: 10.1007/978-3-540-75829-7_13 Dutta-Bergman, M. J. (2004a). Complementarity in consumption of news types across traditional and new media. Journal of Broadcasting & Electronic Media, 48(1), 41–60. doi: 10.1207/s15506878jobem4801_3 Dutta-Bergman, M. J. (2004b). The impact of completeness and Web use motivation on the credibility of e-health information. Journal of Communication, 54(2), 253–269. doi: 10.1111/j.1460-2466.2004.tb02627.x Escoffery, C., Miner, K. R., Adame, D. D., Butler, S., McCormick, L., & Mendell, E. (2005). Internet use for health information among college students. Journal of American College Health, 53(4), 183–188. doi: 10.3200/JACH.53.4.183-188 Fazio, R. H. (1990). Multiple processes by which attitudes guide behavior: The MODE model as an integrative framework. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 23, pp. 75–109). New York, NY: Academic Press. doi: 10.1016/S0065-2601(08)60318-4 Fazio, R. H. (1995). Attitudes as object-evaluation associations: Determinants, consequences, and correlates of attitude accessibility. In R. E. Petty & J. A. Krosnick (Eds.), Attitude strength: Antecedents and consequences (pp. 247–282). Hillsdale, NJ: Erlbaum. Fazio, R. H. (2001). On the automatic activation of associated evaluations: An overview. Cognition and Emotion, 15(2), 115–141. doi: 10.1080/02699930125908 Fazio, R. H. (2007). Attitudes as object-evaluation associations of varying strength. Social Cognition, 25(5), 603–637. doi: 10.1521/soco.2007.25.5.603 Fazio, R. H., & Roskos-Ewoldsen, D. R. (2005). Acting as we feel: When and how attitudes guide behavior. In T. C. Brock & M. C. Green (Eds.), Persuasion: Psychological insights and perspectives (2nd ed., pp. 41–62). Thousand Oaks, CA: SAGE. Feather, N. T. (1962). Cigarette smoking and lung cancer: A study of cognitive dissonance. Australian Journal of Psychology, 14(1), 55–64. doi: 10.1080/00049536208255449 Feather, N. T. (1963). Cognitive dissonance, sensitivity, and evaluation. Journal of Abnormal and Social Psychology, 66(2), 157–163. doi: 10.1037/h0049383 Festinger, L. (1957). A theory of cognitive dissonance. Stanford, CA: Stanford University Press. Festinger, L. (Ed.) (1964). Conflict, decision, and dissonance. Stanford, CA: Stanford University Press. Fox, S. (2011, February 1). Health topics: 80% of Internet users look for health information online. Pew Internet & American Life Project. Retrieved from http://www.pewinternet.org/∼/media//Files/Reports/2011/PIP_Health_Topics.pdf Grace, T. W. (1997). Health problems of college students. Journal of American College Health, 45(6), 243–251. doi: 10.1080/07448481.1997.9936894 Griffin, R. J., Dunwoody, S., & Neuwirth, K. (1999). Proposed model of the relationship of risk information seeking and processing to the development of preventive behaviors. Environmental Research, 80(2), S230–S245. doi: 10.1006/enrs.1998.3940 Hastall, M. R., & Knobloch-Westerwick, S. (2013). Severity, efficacy, and evidence type as determinants of health message exposure. Health Communication, 28(4), 378–388. doi: 10.1080/10410236.2012.690175 Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

823



Hornik, R. C. (Ed.) (2002). Public health communication: Evidence for behavior change. Mahwah, NJ: Erlbaum. Hu, Y., & Sundar, S. S. (2010). Effects of online health sources on credibility and behavioral intentions. Communication Research, 37(1), 105–132. doi: 10.1177/0093650209351512 Hwang, Y. (2010). Selective exposure and selective perception of anti-tobacco campaign messages: The impacts of campaign exposure on selective perception. Health Communication, 25(2), 182–190. doi: 10.1080/10410230903474027 Johnson, J. D., & Meischke, H. (1993). A comprehensive model of cancer-related information seeking applied to magazines. Human Communication Research, 19(3), 343–367. doi: 10.1111/j.1468-2958.1993.tb00305.x Kessels, L. T. E., Ruiter, R. A. C., & Jansma, B. M. (2010). Increased attention but more efficient disengagement: Neuroscientific evidence for defensive processing of threatening health information. Health Psychology, 29(4), 346–354. doi: 10.1037/a0019372 Kleinhesselink, R. R., & Edwards, R. E. (1975). Seeking and avoiding belief-discrepant information as a function of its perceived refutability. Journal of Personality and Social Psychology, 31(5), 787–790. doi: 10.1037/h0076702 Knobloch-Westerwick, S. (in press). Choice and preference in media use: Advances in selective-exposure research. New York, NY: Routledge/Taylor and Francis. Knobloch-Westerwick, S., & Crane, J. (2012). A losing battle: Effects of prolonged exposure to thin-ideal images on dieting and body satisfaction. Communication Research, 39(1), 79–102. doi: 10.1177/0093650211400596 Knobloch-Westerwick, S., & Kleinman, S. B. (2012). Preelection selective exposure: Confirmation bias versus informational utility. Communication Research, 39(2), 170–193. doi: 10.1177/0093650211400597 Knobloch-Westerwick, S., & Sarge, M. A. (in press). Impacts of exemplification and efficacy as characteristics of an online weight loss message on selective exposure and subsequent weight loss behavior. Communication Research. doi: 10.1177/0093650213478440 Liberman, A., & Chaiken, S. (1992). Defensive processing of personally relevant health messages. Personality and Social Psychology Bulletin, 18(6), 669–679. doi: 10.1177/0146167292186002 Lord, R. G., Diefendorff, J. M., Schmidt, A. M., & Hall, R. J. (2010). Self-regulation at work. Annual Review of Psychology, 61, 543–568. doi: 10.1146/annurev.psych.093008.100314 Lowin, A. (1967). Approach and avoidance: Alternate modes of selective exposure to information. Journal of Personality and Social Psychology, 6(1), 1–9. doi: 10.1037/h0024531 Lowin, A. (1969). Further evidence for an approach-avoidance interpretation of selective exposure. Journal of Experimental Social Psychology, 5(3), 265–271. doi: 10.1016/0022-1031(69)90052-3 Metzger, M. J., Flanagin, A. J., & Medders, R. B. (2010). Social and heuristic approaches to credibility evaluation online. Journal of Communication, 60(3), 413–439. doi: 10.1111/j.1460-2466.2010.01488.x Morris, D. S., Rooney, M. P., Wray, R. J., & Kreuter, M. W. (2009). Measuring exposure to health messages in community-based intervention studies: A systematic review of current practices. Health Education & Behavior, 36(6), 979–998. doi: 10.1177/1090198108330001 Pease, M. E., Brannon, L. A., & Pilling, V. K. (2006). Increasing selective exposure to health messages by targeting person versus behavior schemas. Health Communication, 19(3), 231–240. doi: 10.1207/s15327027hc1903_5 824




Pornpitakpan, C. (2004). The persuasiveness of source credibility: A critical review of five decades’ evidence. Journal of Applied Social Psychology, 34(2), 243–281. doi: 10.1111/j.1559-1816.2004.tb02547.x Prior, M. (2009). Improving media effects research through better measurement of news exposure. The Journal of Politics, 71(3), 893–908. doi: 10.1017/S0022381609090781 Prochaska, J. O., DiClemente, C. C., & Norcross, J. C. (1992). In search of how people change: Applications to addictive behaviors. American Psychologist, 47(9), 1102–1114. doi: 10.1037/0003-066X.47.9.1102 Rains, S. A. (2007). Perceptions of traditional information sources and use of the World Wide Web to seek health information: Findings from the health information national trends survey. Journal of Health Communication, 12(7), 667–680. doi: 10.1080/10810730701619992 Rimal, R. N., & Real, K. (2003). Perceived risk and efficacy beliefs as motivators of change: Use of the risk perception attitude (RPA) framework to understand health behaviors. Human Communication Research, 29(3), 370–399. doi: 10.1111/j.1468-2958.2003.tb00844.x Rothman, A. J., Baldwin, A. S., Hertel, A. W., & Fugelstad, P. T. (2011). Self-regulation and behavior change: Disentangling behavioral initiation and behavioral maintenance. In K. D. Vohs & R. F. Baumeister (Eds.), Handbook of self-regulation: Research, theory, and applications (2nd ed., pp. 106–122). New York, NY: Guilford Press. Ruppel, E. K., & Rains, S. A. (2012). Information sources and the health information-seeking process: An application and extension of channel complementarity theory. Communication Monographs, 79(3), 385–405. doi: 10.1080/03637751.2012.697627 Shah, J. Y., & Kruglanski, A. W. (2003). When opportunity knocks: Bottom-up priming of goals by means and its effects on self-regulation. Journal of Personality and Social Psychology, 84(6), 1109–1122. doi: 10.1037/0022-3514.84.6.1109 Slater, M. D. (2004). Operationalizing and analyzing exposure: The foundation of media effects research. Journalism & Mass Communication Quarterly, 81(1), 168–183. doi: 10.1177/107769900408100112 Tian, Y., & Robinson, J. D. (2008). Media use and health information seeking: An empirical test of complementarity theory. Health Communication, 23(2), 184–190. doi: 10.1080/10410230801968260 Turner, M. M., Rimal, R. N., Morrison, D., & Kim, H. (2006). The role of anxiety in seeking and retaining risk information: Testing the risk perception attitude framework in two studies. Human Communication Research, 32(2), 130–156. doi: 10.1111/j.1468-2958.2006.00006.x Turner, M. M., Yao, S., Baker, R., Goodman, J., & Materese, S. A. (2010). Do lay people prepare both sides of an argument? The effects of confidence, forewarning, and expected interaction on seeking out counter-attitudinal information. Argumentation & Advocacy, 46(4), 226–239. Weaver, J. B., III, Mays, D., Weaver, S. S., Hopkins, G. L., Ero˘glu, D., & Bernhardt, J. M. (2010). Health information-seeking behaviors, health indicators, and health risks. American Journal of Public Health, 100(8), 1520–1525. doi: 10.2105/AJPH.2009.180521 Weinstein, N. D. (1979). Seeking reassuring or threatening information about environmental cancer. Journal of Behavioral Medicine, 2(2), 125–139. doi: 10.1007/BF00846662 Weinstein, N. D. (1985). Reactions to life-style warnings: Coffee and cancer. Health Education & Behavior, 12(2), 129–134. doi: 10.1177/109019818501200201 Journal of Communication 63 (2013) 807–829 © 2013 International Communication Association

825



Weinstein, N. D. (1988). The precaution adoption process. Health Psychology, 7(4), 355–386. doi: 10.1037/0278-6133.7.4.355 Willis, L., & Knobloch-Westerwick, S. (in press). Weighing women down: Messages on weight loss and body shaping in editorial content in popular women’s health and fitness magazines. Health Communication. doi: 10.1080/10410236.2012.755602 Wolfram, D., Wang, P., & Zhang, J. (2009). Identifying Web search session patterns using cluster analysis: A comparison of three search environments. Journal of the American Society for Information Science and Technology, 60(5), 896–910. doi: 10.1002/asi.21034 Ye, Y. (2010). A path analysis on correlates of consumer trust in online health information: Evidence from the health information national trends survey. Journal of Health Communication, 15(Suppl. 3), 200–215. doi: 10.1080/10810730.2010.522687

Appendix A: Pilot and Pretests Pilot test of topics

The topics for the presented messages were selected based on a pilot test of seven topics. Forty-one participants (51% female) drawn from the same population responded to ‘‘How relevant are the following health issues to you personally?’’ on a 7-point scale (1 = not relevant at all, 7 = extremely relevant) for seven topics. The four chosen topics all received average ratings above the scale midpoint of 4 and were thus of moderate to high relevance in the participants’ perspective. The three topics with lower ratings (sun exposure, cancer risk from cell phones, and red meat consumption) were not included in the main experiment. Pretest of sources

The names and URLs of a range of health websites were pretested to identify sources that were perceived as having high and low credibility. A separate sample of participants (N = 12, 50% female, M age = 21.67, SDage = 0.98), drawn from the same population as the main study, were presented with the names and URLs of 32 health websites and asked to rate how credible they would expect each website to be, on a 7-point scale ranging from not at all to very much. The eight highest scoring and eight lowest scoring sources were significantly distinct groups, and were selected for inclusion in the study (see Appendix B). Pretest of search results previews

Participants in the pretest sample (N = 12) were also asked to rate the headlines and leads as stimuli to be presented as search results previews in the main study. These previews were rated, using 7-point scales from not at all to very much, on the dimensions of credibility, interest, and support for the topic (e.g., ‘‘supports organic foods’’). The selected previews are shown in Appendix C. Within a topic, articles did not vary by credibility or interest. However, the manipulation of issue stance was successful, with significant differences between supportive and oppositional articles for each health topic (see Appendix D). 826




Appendix B: Perceived Credibility Ratings of Selected Sources for Health Articles Credibility Source High-Credibility Sources Centers for Disease Control—www.cdc.gov National Institutes of Health—www.nih.gov Dept. of Health & Human Services—www.hhs.gov U.S. Food & Drug Administration—www.fda.gov American Public Health Association—www.apha.org American College Health Association—www.acha.org American Dietetic Association—www.eatright.org American Council on Health & Science—www.acsh.org Low-Credibility Sources Overseas-Foreign Pharmacy—www.overseas-foreign-pharmacy.com Health Is Wealth—www.healthiswealthupdate.blogspot.com Shirley’s Wellness Café —www.shirleys-wellness-cafe.com John’s Health Blog—www.johnshealthblog.com My Health Blog—www.myhealth.spot.bz Teenage Health Freak—www.teenagehealthfreak.wordpress.com Nicola Quinn’s Health Blog—www.nicolaquinn.blogspot.com The Magic Pill—www.themagicpill.com

M (1–7)

SD

6.58a 6.42a 6.25a 6.17a 5.75a 5.75a 5.25a 4.83a

0.79 0.90 1.22 1.99 1.66 1.29 1.82 1.99

2.83b 2.67b 2.25bc 2.08bc 2.00bc 2.00bc 1.67bc 1.67c

1.70 1.15 1.22 1.16 1.21 1.04 1.23 0.98

Note: N = 12. Means with different lower case letters differ at p < .01.

Appendix C: Headlines and Leads for Supportive and Oppositional Health Articles Topic Organic Food

Support

Oppose

Organic Not Nutritionally Better: There is no evidence that organic foods are nutritionally superior to conventionally produced food, according to a study in The American Journal of Clinical Nutrition. Organics Rich in Nutrients: Evidence The Organic Food Fad: In 1952, Martin Gardner characterized finds that organic crops contain organic as a food fad without increased nutrients. Analysis of scientific justification. Sixty organic tomatoes, apples, and years later, science shows that peaches revealed greater organic is expensive but not concentrations of vitamin C, healthier.* polyphenols, betacarotene, and flavonoids.

Organic Food: Fewer Pesticides: A detailed scientific analysis of organically grown fruits and vegetables shows that they contain a third as many pesticide residues as conventionally grown foods do.


827



Appendix C: Continued Topic

Support

Oppose

Coffee

Coffee Benefits Equal Vegetables: The regular drinking of coffee is likely to contribute as many health-giving antioxidants to a person’s diet as fruit and vegetables, new research results suggest. Guilt-Free Pleasure of Coffee: With coffee shops seemingly on every corner, and a continued increase in American coffee consumption, the news about coffee’s effects on health is surprisingly good. Eating Vegetables: Makes Big Difference: If Americans want to be healthy, fruits and vegetables are their best friends. New scientific reports demonstrate the science behind the value of eating more. Nutrition: Can’t Beat Plant-Based: Those in the know consider fruits and vegetables the healthiest foods around. A diet rich in fresh produce can make people happier, healthier, and longer-lived. Exercise Supports Health, Longevity: Fitness can slow or reverse many effects of aging. Many people are realizing that their medical fate lies in their commitment to an exercise routine.

Coffee Is Addictive, Mind-Altering: Caffeine addicts may try their best to give up their coffee habit—but usually are not able to, even when it threatens their very well-being.

Fruits and Vegetables

Exercise

Physical Activity Prevents Diseases: Regular moderate workouts may help fight off colds and flu, reduce the risk of certain cancers and chronic diseases, and slow the process of aging.

Coffee: Real Health Risks: Like smokers and drug users, those who consumer large amounts of caffeine may also suffer from harmful repercussions down the line, a recent study found. Fruit, Vegetable: No Cancer Shield: A major study tracking eating habits of 478,000 Europeans suggests that consuming lots of fruits and vegetables has little if any effect on preventing cancer. Produce Causes Food Illnesses: For consumers who took nutritionists’ advice and began eating more fruits and vegetables, word that this is risking their lives can come as a shock. Heavy Workouts Damage Health: For some Americans, exercise can become something of an obsession, pursued despite physical injuries, damaged relationships, and time stolen from work, family, and social activities. Too Much Exercise Hurts: Exceeding recommended amounts of almost all prescribed health practices, from the rigorous to the most innocuous exercise programs and beneficial diets, poses definite health risks.

Note: *After pretesting, the last sentence was changed (originally: ‘‘Nearly 60 years later, science evidence has not changed at all.’’) to increase negativity. 828




Appendix D: Perceived Stance of Health Messages Support Article Topic Support organic 1 Support organic 2 Oppose organic 1 Oppose organic 2* Support coffee 1 Support coffee 2 Oppose coffee 1 Oppose coffee 2 Support vegetables 1 Support vegetables 2 Oppose vegetables 1 Oppose vegetables 2 Support exercise 1 Support exercise 2 Oppose exercise 1 Oppose exercise 2

Credibility

Interest

M (1–7)

SD

M (1–7)

SD

M (1–7)

SD

5.58a 5.83a 1.83b 3.42b 6.25a 5.83a 2.17b 2.00b 6.33a 6.33a 2.50b 2.25b 6.17a 6.33a 2.67b 2.75b

1.93 1.11 1.19 1.00 1.06 1.27 1.90 1.35 1.15 .98 1.83 1.48 1.75 .98 1.92 1.91

4.58 5.00 4.00 4.33 4.33 4.50 4.75 4.42 4.58 4.92 4.08 3.83 5.08 4.83 3.75 4.08

1.44 .85 1.28 .98 1.44 1.09 1.14 .67 1.44 1.62 1.31 1.80 1.00 .94 1.54 1.44

4.42 4.58 4.83 4.25 5.17 5.00 5.17 5.17 4.92 5.00 4.50 4.42 5.25 5.25 4.67 4.42

1.44 1.08 1.34 1.82 1.03 1.13 1.03 1.11 1.62 1.04 1.51 1.56 1.06 1.22 1.56 1.51

Note: N = 12. Means within a topic with different superscripts differ at p < .01. The lead with an asterisk was changed after pretesting to increase negativity, see Appendix C.

Appendix E: Descriptive Statistics

Measure

Organic Food M (SD)

Coffee M (SD)

Dichotomous standard perception ‘‘Good for Your Health’’ Preexposure 96.9%a 44.5%b a Postexposure 85.1% 40.4%b Standard accessibility Preexposure 1108 (372)a 1305 (511)b Postexposure 1427 (774)a 1364 (743)a Importance 4.61 (1.95)a 3.63 (2.09)b Personal behavior 3.20 (2.01)a 2.15 (1.57)b Standard perception 4.62 (2.27)a 2.49 (1.48)b Standard-behavior discrepancy 1.43 (2.17)a 0.34 (1.66)b Total exposure to related articles (s) 102 (21)a 104 (18)b

Vegetables M (SD)

Exercise M (SD)

99.1%c 96.7%c

98.6%ac 96.0%c

1011 (295)c 1154 (595)b 6.16 (1.32)c 4.13 (1.65)c 5.82 (1.83)c 1.69 (1.93)a 104 (18)b

986 (268)c 1212 (657)b 6.36 (1.21)d 4.73 (1.98)d 5.37 (1.45)d 0.65 (2.16)b 106 (19)c

Note: Means and percentages within a row with different superscripts differ at p < .05.


829