This article was downloaded by: [Ecole Hautes Etudes Commer-Montreal] On: 05 September 2015, At: 14:03 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: 5 Howick Place, London, SW1P 1WG

Health Communication Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/hhth20

Antecedents to Psychological Reactance: The Impact of Threat, Message Frame, and Choice a

Lijiang Shen a

Department of Communication Studies, University of Georgia Published online: 25 Sep 2014.

Click for updates To cite this article: Lijiang Shen (2015) Antecedents to Psychological Reactance: The Impact of Threat, Message Frame, and Choice, Health Communication, 30:10, 975-985, DOI: 10.1080/10410236.2014.910882 To link to this article: http://dx.doi.org/10.1080/10410236.2014.910882

PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http:// www.tandfonline.com/page/terms-and-conditions

Health Communication, 30: 975–985, 2015 Copyright © Taylor & Francis Group, LLC ISSN: 1041-0236 print / 1532-7027 online DOI: 10.1080/10410236.2014.910882

Antecedents to Psychological Reactance: The Impact of Threat, Message Frame, and Choice Downloaded by [Ecole Hautes Etudes Commer-Montreal] at 14:03 05 September 2015

Lijiang Shen Department of Communication Studies University of Georgia

A 2 (Threat: high vs. low) × 2 (Frame: gain vs. loss) × 2 (Choice: yes vs. no) × 2 (Behavior: prevention vs. detection) factorial design Web-based experiment (N = 814) was conducted to investigate the impact of threat to freedom, message frame, and behavioral choice as antecedents to psychological reactance. The intertwined model for reactance measure was replicated. Results showed that threat to freedom and the loss frame increased, and the gain frame and choice mitigated psychological reactance. The advantages of choice and the gain frame were most salient when threat was high.

The most prominent progress in communication research on psychological reactance since the review by Burgoon, Alvaro, Grandpre, and Voulodakis (2002) has been the operationalization of the reactance construct such that it can be measured as anger and negative cognition intertwined. The intertwined model of psychological reactance has been replicated in multiple studies (for a review, see Quick, Shen, & Dillard, 2013) and further confirmed by a meta-analysis (Rains, 2013). Knowledge of the nature of psychological reactance sheds light on our understanding of why and when persuasion fails. Some scholars went further and argued that negative cognition and anger intertwined could be considered as a more general form of resistance to persuasion (e.g., Kim, Levine, & Allen, 2013). Scholars also agree that research on psychological reactance has important practical implications for message design in public health communication. Burgoon et al. (2002) argued that the relationship between psychological reactance and message features should be studied if we were to understand the acceptance or rejection of health-related campaign messages. To fully utilize the empirical research findings on reactance in practice would require better understandings and more knowledge of message features that would arouse as well as diminish psychological reactance. Quick et al. (2013) highlighted studies that advance understandings of

Correspondence should be addressed to Lijiang Shen, Associate Professor, Department of Communication Studies, University of Georgia, 604 Caldwell Hall, Athens, GA 30602. E-mail: [email protected]

such message features as the most promising possibility in future research on psychological reactance. So far, most reactance studies have focused on freedomthreatening messages that are intense, explicit, and use concrete language with a clear intent to persuade (see Table 1 of Rains, 2013, for sample texts). Far fewer efforts have been devoted to reactance-reducing message features. One study examined the impact of postscripts that might restore perceived choice options (Miller, Lane, Deatrick, Young, & Potts, 2007). At a more global and macro level, narrative persuasive messages (e.g., Dahlstrom, 2012; Moyer-Gusé & Nabi, 2010) and empathy appeal messages (e.g., Shen, 2010) have been found to overcome psychological reactance. There is a gap in knowledge regarding the micro-level message features that reduce reactance. It is imperative to focus on reactance-arousing message features when the goal of the studies is to explore and assess different measurement models of reactance. Nevertheless, practical implications from such studies on reactance-arousing message features tend to be limited in that they are only proscriptive in nature. That is, findings from such studies only tell us what not to do in message design—for example, we should not use forceful, explicit, or concrete messages and we should not demonstrate clear intent to persuade. For both theoretical and practical reasons, more research on reactance diminishing message features is needed. Studying reactancediminishing message features might allow us to test the measurement model in a more comprehensive nomological network. Findings from such studies would also be able to offer prescriptive guidance for message design. That is, we

Downloaded by [Ecole Hautes Etudes Commer-Montreal] at 14:03 05 September 2015

976

SHEN

would know better what to do to design persuasive messages that won’t activate reactance. Quick et al. (2013) called for more theory-driven research on both reactance-arousing and reactance-diminishing message features. The primary goal of this article is to answer their call and to explore the impact of message features that might elicit as well as reduce psychological reactance. First, hypotheses are derived regarding three message features that are antecedents to psychological reactance, namely, threat to freedom, message framing, and behavioral choice. Second, data from a Web-based experiment are presented. Finally, implications for future research and application of persuasive health communication are discussed.

THE IMPACT OF THREAT TO FREEDOM Freedom The central tenet of the psychological reactance theory is that when individuals’ freedom is restricted or removed, psychological reactance will be activated and individuals will strive to restore the lost freedom. Freedom in psychological reactance theory is not freedom in general terms; it is “not abstract considerations, but concrete behavioral realities” (Brehm & Brehm, 1981, p. 12). Knowledge of and ability to exercise a free behavior are two prerequisites for an individual to have the freedom. Free behaviors are not limited to molar instrumental acts, but also include emotions, attitudes, and any other feeling states of the organism (Wicklund, 1974). Threat to Freedom Given that an individual perceives a specific freedom, any force on the individual that makes it more difficult for him or her to exercise the freedom constitutes a threat (Brehm & Brehm, 1981). The social influence attempt in persuasion, be it to shape, reinforce, or change responses (Miller, 2013), constitutes a threat to freedom. Even mere exposure to information, if it makes it difficult for the individual to make a preferred decision, can constitute a threat (e.g., Miller, 1976). The magnitude of psychologial reactance increases as the following increase: importance of the threatened freedom, and the number and proportion of freedoms threatened. Message Features and Threat to Freedom Threat to freedom is conceptualized as the proximate antecedent of psychological reactance. Manipulation of this consrtuct is primarily achieved through the use of controlling, dogmatic, explicit and vivid language. Public health communication and persuasion in general take place in the realm of nonclose relationships, where, according to politeness theory (Brown & Levinson, 1987), the more

direct a message is, the stronger is the threat to one’s need for autonomy and self-determination (Jenkins & Dragojevic, 2013). Hence, there is consensus from multiple theoretical perspectives that explicit and strong language would threaten individuals’ freedom, which further causes resistance to persuasion. Inherent in persuasive messages is the intent to persuade. Studies on reactance and boomerang attitude changes have provided strong evidence for the main effect of intent to persuade on decreased persuasion (Benoit, 1998; Worchel & Brehm, 1970). Research in other theories also suggests that perceived intent to persuade inhibits persuasion. There is strong evidence indicating that awareness of the intent to persuade from the influence agent leads to resistance to the message, even when it is just a forewarning without the actual message (Compton, 2013; Pfau, 1997). Research in narrative persausion also suggests that reduced awareness of the intent to persuade is one of the key mechanisms that gives narrative messages the persuasive advantage (Bilandzic & Busselle, 2013). Based on both theory and empirical evidence, the following hypothesis was predicted: H1: Threat to freedom is positively associated with psychological reactance. Note that there was nothing novel in this hypothesis. It was supposed to replicate existing empirical findings. On the other hand, it was also necessary. Scholars have emphasized that the relationship between threat to freedom and psychological reactance should be estbalished to validate the reactance measure (e.g., Dillard & Shen, 2005; Quick, 2012; Quick & Stephenson, 2008). Therefore, this hypothesis was also considered as the prerequiste for testing other hypotheses and investigation of message features as antecedents to psychological reactance.

THE IMPACT OF MESSAGE FRAME Psychological reactance has also been used as an explanation for the relative effectiveness of gain versus loss message framing (e.g., Cho & Sands, 2011; Reinhart, Marshall, Feeley, & Tutzauer, 2007). Message framing is the persuasive strategy either to highlight benefits and rewards from compliance with the message, or to present the costs and punishments that come with noncompliance. O’Keefe and Jensen (2006) demonstrated that neither frame was inherently more effective than the other. And there were no framing effects within particular contexts except for one particular beahvior: There was no framing effect within prevention beahvior except for dential hygiene behaviors (O’Keefe & Jensen, 2007; see also O’Keefe & Wu, 2012) or within detection behavior except for breast cancer selfexams (O’Keefe & Jensen, 2009). This result is quite the opposite from the framing research in prospect theory, where there is strong evdience in favor of the loss frame (Levin,

Downloaded by [Ecole Hautes Etudes Commer-Montreal] at 14:03 05 September 2015

ANTECEDENTS TO REACTANCE

Scheneider, & Gaeth, 1998). It should be noted that the operationalization of framing in prospect is combined with different levels of uncertainty, which message framing in persuasion does not have. In addition, in persuasive message framing, the expected utility across the two frames might not be equal, as is the case in prospect theory (Brendl, Higgins, & Lemm, 1995; Salovey, Schneider, & Apanovitch, 2002; for a typology of framing, see also Levin et al., 1998). Psychological reactance has been conceptualized as the factor that offsets the negativity bias, which is believed to underly the effectiveness of the loss frame. Operationally, a loss-framed persuasive message overlaps with the threat component in a fear appeal message in that they both highlight negative consequences (i.e., in terms of severity and susceptibility). Although conceptually the negative frame and fear appeals are not the same since the negative frame could active any negative emotion, empirically they have a lot in common. There has been empirical evidence that the loss frame indeed results in stronger fear arousal (e.g., Millar & Millar, 2000; Schneider et al., 2001; Shen & Dillard, 2007). There are several reasons that the negative frame might induce psychological reactance. First, by highlighting negative consequences, the language used in the loss frame tends to be more forceful, intense, and controlling (Cho & Sands, 2011). Second, persuasion intent might be more salient in the loss frame in that it might be perceived as “a command that must be answered, obeyed, or reacted against” (Cho & Sands, 2011, p. 310). Third, given that the loss frame leads to stronger negative emotions, it might be perceived as more manipulative (Witte, 1992). In addition to the boomerang effect from fear appeal messages (e.g., Janis & Feshbach, 1953), there is direct evidence that fear appeal messages lead to stronger psychological reactance (e.g., Shen, 2011). There also has been empirical evidence for the impact of negative frame on psychological reactance, but the evidence has not been as clear. Reinhart et al. (2007) demonstrated evidence for the impact of the loss frame on psychological reactance; however, the Lindsey (2005) measure of psychological reactance was adopted, which has been deemed less appropriate (Quick, 2012). In Cho and Sands (2011), there was evidence that the loss frame produced a greater perceived threat to freedom, yet there was no direct evidence for its impact on psychological reactance as measured by the combination of anger and negative cognition. A third study, Shen and Dillard (2007), did not address psychological reactance directly, but was relevant: There was evidence that the loss frame led to stronger anger arousal; however, there was no information regarding its impact on negative cognition or perceived threat to freedom. In summary, both theory and available empirical evdience suggest that the loss frame might be an antecedent of psycholgoical reactance:

977

H2: The loss frame leads to stronger psychological reactance than the gain frame.

THE IMPACT OF CHOICE Earlier research on psychological reactance primarily focused on the impact of restricted alternatives or forced choices (see a review in Brehm & Brehm, 1981). To a certain degree, the notion of knowledge of a free behavior has been conceptualized in terms of availability of alternatives (Brehm & Sensenig, 1966; Wright, 1986). On the other hand, when an individual loses feelings of choice, learned helplessness occurs and the individual does not perceive any freedom (Wortman & Brehm, 1975). Consequently, there will be no psychological reactance even when a freedom is eliminated or threatened. There has been empirical evidence for the positive impact of choice on behavioral outcomes. Zeinstra, Renes, Koelen, Kok, and de Graaf (2010) demonstrated that offering choices of vegetables significantly improved kids’ subjective meal experience, and lack of choice led to significantly less vegetable consumption among highly reactant children. In a recent meta-analysis, Patall, Cooper, and Robinson (2008) found that providing choices had positive effects on outcomes such as motivation, task performance, perceived competence, satisfaction, and others. Communication scholars have investigated the impact of language features that restrict or enhance choice. Quick and Stephenson (2007) found that threat to choice messages were positively associated with psychological reactance. Miller et al. (2007) examined the impact of postscripts that at least partially restored one’s choice. A restoration postscript tells the recipient, “The choice is yours. You are free to decide for yourself,” while a filler postscript does not contain such an argument. Their results showed that the restoration postscript significantly reduced perceived threat to freedom. However, there was no direct evidence for the impact of such postscripts on psychological reactance itself. Consider that multiple behavioral options might be available to reduce the same health risk; for example, sun safety can be achieved with sunscreen application or wearing protective clothing, although the perceptions of each behavior might be quite different. Providing information on alternative behavioral options might reduce psychological reactance since the threatened freedom is now only a fraction of the total freedom (Wicklund, Slattum, & Solomon, 1970). According to psychological reactance theory, this should hold regardless of language intensity or intent to persuade. The utility in offering a “replacement” behavior in efforts to reduce or eliminate a problem behavior (Johnston, 2006) might just come from the notion of choice and reduction of psychological reactance. Hence, it was predicted that:

978

SHEN

H3: Providing an alternative behavioral option is negatively associated with psychological reactance.

Downloaded by [Ecole Hautes Etudes Commer-Montreal] at 14:03 05 September 2015

THE INTERPLAY OF MESSAGE FEATURES Scholars have suggested that the impact of message features is mediated by individuals’ construal of such features (e.g., O’Keefe, 2003). When individuals are exposed to multiple message features, it is possible that individuals’ perception and processing of such message features might interact with each other. An interaction among these message features might also be viewed as compatible with the theoretical frameworks that suggest the main effects. Available empirical evidence suggested that the impact of other factors on psychological reactance would be magnified when the threat to freedom level was high and/or trait reactance was high, but not when threat to freedom or trait reactance was low (e.g., Bensley & Wu, 1991; floss data, Dillard & Shen, 2005; Karno & Longabaugh, 2005; Zeinstra et al., 2010). Hence, the following two sets of interaction effects were predicted: H4: Choice and threat to freedom interact in such a way that: H4a: The no choice condition induces stronger reactance than the choice condition when threat to freedom is high. H4b: There is no such effect from choice on reactance when the threat to freedom is low. H5: Message frame and threat to freedom interact in such a way that: H5a: The loss frame induces stronger reactance than the gain frame when threat to freedom is high. H5b: There is no message framing effect on reactance when the threat to freedom is low.

The construct of behavior type was used as a controlled covariate in data analyses. The topic of skin cancer was selected because (a) it was relevant and involving to the sample in this study (i.e., college students), and (b) there were multiple options available either to prevent the risk or to detect early symptoms of skin cancer. The messages differed only in the recommendation component, where the manipulations occurred. High threat to freedom messages used strong, dogmatic, and controlling language with a clear intent to persuade—for example, “the evidence is extremely clear” and “you must do behavior X.” Low threat to freedom messages used mild and tentative language with a less explicit intent to persuade—for example, “there is evidence” and “you might want to consider doing behavior X.” Messages in the gain frame highlighted the benefits and rewards associated with compliance with the messages, such as healthier and better looking skin, avoiding such symptoms as irritation, blisters, swelling, skin damage, and lowered risk of skin cancer. Messages in the loss frame emphasized the negative consequences from noncompliance, such as suffering from the symptoms of irritation, blisters, peeling skin and swelling, damaged skin and photo aging, and increased risk of skin cancer. The choice messages discussed two behavioral alternatives. For prevention behavior, the options of sunscreen use and wearing protective clothing were offered. The recommendation was either to apply sunscreen regularly or to wear protective clothing. For detection behavior, the options of self-examination and making an appointment with a dermatologist for a checkup were presented. The recommendation was to either perform self-examination or make an appointment with a dermatologist for a checkup. The no choice messages only recommended sunscreen use for prevention behavior and self-examination for detection behavior. These four behavioral recommendations varied in message frame and threat to freedom, resulting in 16 versions of messages corresponding to the conditions in the experimental design.

METHOD Overview Experimental design and stimuli messages. The study was a 2 (threat to freedom: high vs. low) × 2 (frame: gain vs. loss) × 2 (behavioral choice: yes vs. no) × 2 (behavior: prevention vs. detection) factorial design. The stimuli were 16 messages on skin cancer, designed in a threatrecommendation format. Each one first described the risks of skin cancer, and then recommended a behavior that either would prevent it (sunscreen application and wearing protective clothing) or would detect early symptoms of skin cancer (self-examination and getting tested for skin cancer). The contrast of prevention versus detection behaviors was not of theoretical interest in this study. It was included because it might moderate message framing effect (e.g., O’Keefe & Jensen, 2007, 2009; Salovey et al., 2002). Inclusion of more than one type of behavior also enhanced external validity.

Participants Participants were 815 students recruited from undergraduate classes in communication at the University of Georgia. Participation in the study either fulfilled their course requirement or earned them a small portion of extra credit. The participants ranged in age from 18 to 32 years (M = 19.54, SD 1.63), with 77.1% describing themselves as white/Caucasian, 7.8% as of African descent, 5.9% as of Asian descent, 2.8% as of Hispanic descent, and 6.3% as other. Fifty-nine percent reported their sex as female and 41% as male. Actual sample size varied in data analyses due to missing values. Procedure Data collection took place online. The 16 stimuli messages and measurement instruments were set up on

Downloaded by [Ecole Hautes Etudes Commer-Montreal] at 14:03 05 September 2015

ANTECEDENTS TO REACTANCE

SurveyMonkey, resulting in 16 versions of the survey. A separate webpage was set up as the informed consent page. Clicking the “consent” button on that page activated an automated algorithm that randomly assigned the participants to one of the 16 conditions. The participants first reported their previous skin cancer-related behaviors including the frequency of sun bathing, the use of indoor tanning, sunscreen use, and self-examination. Then they read the threat component of the skin cancer message, before providing affective responses, and then listing whatever came to their mind when they were reading the message. Next, they responded to questions assessing their perceived risk of skin cancer. Afterward they read the recommendation component of the message before completing the affective response and thought-listing tasks one more time. Next, they responded to questions assessing perceived effectiveness of the message, before they reported their attitude toward message advocacy. The entire session was about 35–40 minutes. Not all data were reported here. Measures Induction checks. Perceived threat to freedom was measured by four 1–5 point (1 = strongly disagree, 5 = strongly agree) Likert-type scale items: “The message threatened my freedom to choose,” “The message tried to make a decision for me,” “The message tried to manipulate me,” and “The message tried to pressure me.” The alpha reliability of the scale was .84. The induction check for message framing manipulation was measured by four 5-point semantic differential items. The participants were asked to make a judgment about the emphasis of the end state presented in the message on the following word pairs: costs/benefits; loss/gain; advantages/disadvantages; and negative/positive outcomes (α = .90). Anger. Anger was measured using four items that had been validated in previous studies (Dillard, Plotnick, Godbold, Freimuth, & Edgar, 1996): irritated, angry, annoyed, and aggravated (α = .93). The 5-point response scale was anchored at 0 = none of this feeling and 4 = a great deal of this feeling. It should be noted that 42.3% of the participants reported “none of this feeling.” Cognitive response. Participants were asked to list whatever was in their minds when they finished reading the action component of the message. Following Dillard and Shen (2005), the resulting data were coded in a fourstep sequence. In Step 1, the coders segmented the data into psychological thought units. A thought unit is the minimum meaningful utterance having a beginning and an end. It is typically a simple sentence or independent clause (Hatfield & Weider-Hatfiled, 1978). Gutzkow’s U was around .03 across the coder pairs for unitization. Affective responses were identified and removed in Step 2. Coders

979

relied on a list of feeling terms compiled by Shaver, Schartz, Kirson, and O’Connor (1987), supplemented by affective terms that emerged from the data. A unit was classified as affective whenever those words appeared and cognitive otherwise. In Step 3, coders evaluated whether or not the cognitive responses were relevant to the message (topic, issue, and task). The purpose of this step was to eliminate irrelevant cognitions and thereby reduce the level of noise in the data. In Step 4, the remaining data were coded either as (a) supportive thoughts, (b) neutral thoughts, or (c) negative thoughts. Supportive thoughts were defined as responses that expressed agreement with the message, self-identification and positive evaluation toward the message, the message source, or the advocacy, intention to comply with the advocacy in the message, and so on. Negative thoughts were defined as responses that refuted or expressed disagreement with the message, disliking of the message source, negative intention to comply with the advocacy, intention to engage in the risk behavior, and so on. Neutral thoughts were defined as nonevaluative responses to the message. Intercoder reliabilities were established with 10% of the data coded by three coders working in three-way pairs before the coders went on to code the remainder of the data individually. Krippendorf’s alpha (Krippendorf, 2004) was calculated with the Concord package in the open-source software R for Steps 2–4: .94 for Step 2, .87 for Step 3, and .82 for Step 4. Only the negative cognitions were used in subsequent data analyses (M = 1.79, SD = 1.39). It should be noted that 20.4% of the participants reported no negative thoughts: Some did not list any thought at all, and others generated thoughts that were not evaluative. Perceived effectiveness. Perceived effectiveness was considered as a proxy measure and an antecedent of actual effectiveness of persuasive messages (Dillard, Shen, & Vail, 2007). Perceived effectiveness was measured by seven 5point semantic differential items adopted from Dillard et al. The items were the following words and their antonyms: convincing, believable, sensible, biased, distorted, fair, and balanced. The items were averaged into a composite score, with higher scores indexing better effectiveness (α = .89). Attitude. Given that two behaviors were recommended in the choice condition, evaluation of a single behavior was not optimal to assess persuasive outcomes. Instead, attitude toward “the message advocacy” was measured. This measure of attitude was more in line with yielding in McGuire’s (1968) information processing model. Attitude toward the message advocacy was measured by four 5-point Likert items: “I agree with what the message recommends,” “I support what the message advocates,” “I am in favor of the position in the message,” and “I endorse the claims made in the message” (α = .87).

980

SHEN

RESULTS

Downloaded by [Ecole Hautes Etudes Commer-Montreal] at 14:03 05 September 2015

Manipulation Check To check the manipulations of threat to freedom and message framing, a multivariate generalized linear model (GLM) was estimated to predict perceived threat to freedom and perceived end state, with the three experimental factors (threat to freedom, message frame, and choice) as predictors, and behavior type, past behaviors, gender, and age as covariates. Pairwise comparisons were carried out via the Bonferroni method to adjust for possible type I error inflation due to multiple tests. With these parameters and a sample size of N = 814, assuming a = .05 (two-tailed), the statistical power to detect an effect size equivalent to r = .10 exceeded .81. No manipulation check was performed for behavioral choice because it was an intrinsic message feature (O’Keefe, 2003). There was a significant main effect of threat manipulation: F(2, 787) =42.37, p < .001, η2 = .098. The main effect of frame manipulation was also significant: F(2, 787) = 35.94, p < .001, η2 = .084. No other terms emerged as significant. Univariate effects were examined to assess the manipulations of threat to freedom and message framing, respectively. Perceived threat to freedom. There was a significant main effect of threat manipulation: F(1, 788) = 81.99, p

Antecedents to Psychological Reactance: The Impact of Threat, Message Frame, and Choice.

A 2 (Threat: high vs. low) × 2 (Frame: gain vs. loss) × 2 (Choice: yes vs. no) × 2 (Behavior: prevention vs. detection) factorial design Web-based exp...
174KB Sizes 0 Downloads 5 Views