The mechanism of self-persuasion:
A study on the effect of using questions (closed-ended vs. open-ended)
instead of statements on the psychological reactance and persuasive
effects of health warnings
Master’s thesis
Graduate School of Communication Research Master Communication Science
Word count: 8692 Suzanne Hinke (10854126) Supervisor: Prof. Dr. Bas van den Putte
Abstract
Excessive alcohol consumption is related to several health problems and is currently labelled as one of the leading risk factors for mortality in Europe and therefore an important issue to tackle. Health campaigns targeted towards responsible alcohol consumption usually lack effects due to defensive reactance towards these campaigns. Therefore, this study explores a potential new technique to circumvent this state of reactance by activating a process of self-persuasion, which is induced by formulating a health warning as a question instead of a statement. Following a one-factor
(statements vs. closed-ended questions vs. open-ended question) between-subjects design, the current study aims to test whether a question results in less reactance as opposed to a statement and what role the type of question (closed-ended vs. open-ended) plays in this. Additionally, expectations are that level of involvement and need for cognition (NfC) influence the level of reactance experienced. Results portray that formulating a health warning as a question instead of a statement has no influence on reactance and persuasive effects of the message and neither does the type of question. Lower defensive reactance does however lead to more persuasive effects of the message and highly involved individuals report more reactance as opposed to low involved individuals. No effects of NfC are found. In general, the current study provides insight into the mechanisms through which self-persuasion works and the importance of taking into account reactance when structuring a health message. Finally, suggestions for future research are given.
Keywords: self-persuasion, psychological reactance, questions, alcohol consumption, involvement, need for cognition
Introduction
A lot of research has been directed towards studying the effects of new techniques and methods with the aim of constructing interventions to improve health campaigns targeted to reduce alcohol consumption (White et al., 2010). However, the
effectiveness of these campaigns seems to remain very small and often little to no actual behaviour change is established (Hornik, Jacobsohn, Orwin, Piesse & Kalton, 2008). This lack of effect is problematic because excessive alcohol consumption can lead to several health related problems and can even be labelled as one of the major leading risk factors for mortality in Europe (World Health Organization, 2014). The Health Council of the Netherlands (2015) has recently published new alcohol consumption guidelines, in which the maximum standard glasses of alcohol
consumption has been set to one for both men and women. Because excessive alcohol consumption, also referred to as ‘binge drinking’, is a prevalent problem in the
Netherlands (Hanewinkel et al., 2012), it is important to make people aware of the advised guidelines for responsible drinking (Gezondheidsraad, 2015).
The absence of large effects of current campaigns can be better understood by taking into account the resistance people might have against health messages. One way to persuade people to eventually change their behaviour is by presenting them with arguments or information regarding the potential risks that are associated with the specific behaviour, in this case excessive alcohol consumption. However, persuading people with arguments by showing the negative consequences of certain behaviour is not always effective because these arguments can trigger reactance, which can result in the target group to decrease the importance of the information or to reject the information as a whole (Brehm, 1966; Dillard & Shen, 2005).
consequences and risks of excessive alcohol consumption can loose its persuasive power, will not lead to the desired behaviour change and might even result in a so called boomerang-effect in which an increase of the risk behaviour could be observed (Miller, Burgoon, Grandpre & Alvaro, 2006; Ringold, 2002). Therefore, it is
necessary to look for potential alternative techniques to circumvent this state of psychological reactance.
One of the new methods that can function as a useful tool to overcome
psychological reactance is the concept of self-persuasion. Self-persuasion involves an individual persuading him or herself internally by thinking of arguments against certain behaviour instead of being persuaded by arguments that are conveyed within a message (Mussweiler & Neumann, 2000). Letting an individual think of arguments him or herself instead of providing someone with arguments can activate
self-persuasion. The positive effects of thinking of arguments internally are explained by the fact that external provided arguments are often evaluated more critically and therefore easier rejected than arguments that are generated internally (Mussweiler & Neumann, 2000). In order to stimulate the process of self-persuasion, a message that is constructed in the form of a question instead of a statement could trigger someone to internally think of arguments (Godin et al., 2012), which in turn reduces
psychological reactance due to the fact that a person does not feel as if the message is forcing an opinion upon someone and hereby threatening the freedom (Brehm, 1966).
Self-persuasion has already proven to be an effective technique in diminishing defensive responses towards health warnings and establishing more accurate risk perceptions in the field of smoking cessation (e.g., Müller et al., 2009, 2016). To date however, little to no research has incorporated this technique within health warnings targeted towards excessive alcohol consumption. Therefore, within my study, it will
be researched whether self-persuasion can function as an effective method to reduce psychological reactance towards health warnings regarding alcohol consumption and hereby increasing the persuasive effects as intended by the message.
In addition to the concept of self-persuasion, the effect of the type of question that is asked will be studied. It could be argued that open-ended questions are more effective compared to closed-ended questions due to the fact that open-ended questions leave more room for interpretation and thus might result in less
psychological reactance towards the message than closed-ended questions because people prefer to have options over limited choice (Grandpre, Alvaro, Burgoon, Miller & Hall, 2003; Schwarz & Oyserman, 2001). Finally, it will be studied whether higher levels of need for cognition (NfC) and more involvement with the topic of alcohol influences the effect of a question over a statement, based on the assumption that more involvement and higher NfC lead to more elaboration and thus higher chances of reactance (Cacioppo & Petty, 1982; Petty & Cacioppo, 1986).
To conclude, my study has a high scientific relevance in the field of
communication science because it contributes to the existing knowledge regarding the activation of self-persuasion within the field of smoking and expands this knowledge to another behaviour, namely alcohol consumption. This study could furthermore add to the current knowledge by researching the potential mediating effect of
psychological reactance. This mediating role has to date not yet been measured but could provide insightful information regarding through which mechanisms self-persuasion works. Finally, for the first time a distinction is made between the effect of open-ended versus closed-ended questions. In contrast to other established methods to reduce reactance towards messages, the concept of using (open-ended) questions to increase the persuasive effects of a health message by reducing psychological
reactance can easily be implemented in practice. Governmental campaigns or
persuasive messages from organizational institutions could simply change the format of their campaign material by including questions instead of statements.
The aim of this research is to gain scientific insight into the activation of self-persuasion through psychological reactance and the effects of open-ended versus closed-ended questions in effective health warning messages regarding alcohol consumption as well as the moderating role of involvement and NfC. This research aim leads to the following research questions.
RQ: Does a health warning formulated as a question (closed-ended versus
open-ended) result in more persuasive effects of the health warning via the state of
psychological reactance compared to a health warning formulated as a statement and what role does involvement with the topic and NfC play in this?
Theoretical framework
This theoretical framework will start with an overview of the role that psychological reactance plays in the effectiveness of health messages and will consequently
introduce the concept of self-persuasion, which could function as an important technique to circumvent this psychological reactance, in order to establish the desired persuasive effects of a health message.
Psychological reactance
Brehm first proposed the theory of psychological reactance in 1966 and defined it as “the motivational state that is hypothesized to occur when a freedom is eliminated or threatened with elimination” (Brehm & Brehm, 1981, p. 37). To this notion, there are four relevant elements that should be considered: Freedom, threat to freedom,
reactance and restoration of freedom (Dillard & Shen, 2005). The theory of
psychological reactance states that if at any point someone feels an external influence threatening an individual’s freedom to perform certain behaviour, reactance can occur towards this external influence in order to restore the freedom. The regaining of freedom can be done directly (e.g., by performing more of the behaviour that was threatened) or indirectly by for example ignoring the threat, derogating the source of the threat and or by performing another behaviour in order to regain the feeling of control over ones own behaviour (Dillard & Shen, 2005).
Health campaigns consist of persuasive messages that are designed to influence the receiver to perform certain behaviour or to not perform an undesired behaviour. In the case of alcohol consumption, a health message that tries to persuade someone into drinking less alcohol can be perceived as a threat to someone’s freedom because such a message tries to persuade someone into not performing a certain behaviour one might want to perform. Therefore, the messages are likely to be
indirectly or directly rejected in order to restore the original freedom (Dillard & Shen, 2005).
Because of these unwanted effects, several techniques have been researched that could potentially contribute to decreasing this psychological reactance such as for example self-affirmation (Steele, 1988). Self-affirmation refers to the process
whereby someone is reflecting upon personal values, attributes or social relations in order to increase one’s feeling of self-integrity, which is found to diminish the defensiveness towards threatening messages (Harris, Mayle, Mabbott & Napper, 2007; Harris & Epton, 2009). One major downside of this technique is the difficulty of deducing this process of self-affirmation in practice: Getting someone to reflect on several positive attributes before being exposed to a health warning or message is
challenging to establish. However, another technique that has proven to be effective in reducing psychological reactance and that could perhaps be more easily
implemented in practice is the concept of self-persuasion (Müller et al., 2016).
Elaboration likelihood model and self-persuasion
The role of self-persuasion in persuasion attempts is emphasized in the elaboration likelihood model (Petty & Cacioppo, 1986). The elaboration likelihood model (ELM) proposes two routes of information processing: The central route (activated when elaboration is high) and the peripheral route (activated when elaboration is low). According to Petty and Cacioppo (1986) the key variable in this model is
involvement: When someone is highly involved with a message he or she is
processing the information more elaborate (e.g., elaborating on strength of arguments, judging accuracy of information) and therefore persuasive effects will be established through the central route of the ELM. In contrast, someone who is less involved will not elaborate on the content of the message and potential persuasion effects are likely to be caused by factors other than the message such as peripheral cues (e.g., colours, images, sounds). Attitude change that occurs via the central route of the ELM is more stable and less subject to counterarguments and therefore also more predictive of behaviour, as opposed to attitude change via the peripheral route (Petty & Cacioppo, 1986).
The concept of self-persuasion can be accommodated in the central route of the ELM because self-persuasion triggers a process of higher elaboration with the message (Mussweiler & Neumann, 2000). Self-persuasion can be explained as the process whereby someone persuades him or herself by thinking of arguments instead of being persuaded by a message, which is found to be more effective than being persuaded by an external source (Blankenship & Craig, 2006; Maio & Thomas,
2007). Because someone is internally thinking of arguments, the central route of the ELM is activated and the attitude change that is established is more stable (Petty & Cacioppo, 1986). Furthermore, thinking of arguments against or in favour of certain behaviour is more effective than receiving arguments in a message, because people process information that comes from an external source more critically compared to arguments that are generated internally (Mussweiler & Neumann, 2000). People are more prone to correct for information provided by an external source because such a source is often less trusted and the information is more likely to be perceived as less accurate than self-generated arguments, which results in the external information to be adjusted according to personal views (Hoch & Deighton, 1989; Mussweiler & Neumann, 2000). It was furthermore found that people think more positively about their own arguments and evaluate self-generated arguments more favourably compared to other type of arguments (Baldwin, Rothman, Vander Weg &
Christensen, 2012). Additional explanations regarding the way self-persuasion can change a person’s attitudes can be found in studies regarding self-perception theory (Bem, 1965, 1967). According to this theory, an individual’s evaluation of attitudes and beliefs can be affected by the own behaviour one observes. Thus, if a binge drinker observes him or herself arguing against this behaviour, these arguments are evaluated as more convincing than arguments provided against binge drinking by an external source.
The effect of self-persuasion in health communication
In the field of health communication, self-persuasion could be considered a useful tool to increase the effectiveness of health campaigns. The process of self-persuasion can be triggered by reformulating statements into questions and hereby changing the format of the message (Godin et al., 2012). Several empirical studies in the field of
health communication have found evidence for the persuasive effect of reframing health warnings into questions instead of statements. A study by Glock, Müller and Ritter (2013) investigated whether constructing statements regarding the dangers of smoking into questions could result in higher risk perceptions. Their results showed that participants who were exposed to a question instead of a statement reported higher smoking-related risk perceptions than the participants who were portrayed to statements. The authors explain this finding by the lower defensive reactance experienced by these participants (as a result of being exposed to questions as opposed to the statements) that probably mediated the effect the health warnings had on their risk perceptions, but did not include a measure of reactance to test this explanation (Glock et al., 2013). Besides higher risk perceptions, it has been found that self-persuasion could also influence actual behaviour change regarding smoking cessation (Müller et al., 2009, 2016). Participants in the study by Müller et al. (2016) were shown a movie in which either questions or statements regarding the dangers of smoking were included. Results showed that participants who were exposed to the smoking-related questions waited longer before lighting up cigarette than participants who were exposed to the smoking-related statement stimuli.
Despite the promising persuasive effects of formulating health warnings into questions as opposed to statements, explained by reduced defensive reactance (Mussweiler & Neumann, 2000), none of these studies have included a measure of defensive reactance to test whether the effects that were found can indeed be contributed to lower defensive reactance as a result of generated self-persuasion (Glock et al., 2013; Müller et al., 2009, 2016). Within my study, this expectation will be tested and leads to the first hypothesis.
H1: A health warning that is formulated as a question will lead to less defensive reactance towards this message compared to a health warning formulated as a statement, which results in more persuasive effects of the message
Type of question
To my knowledge only one study has investigated the effect of self-persuasion with regard to excessive alcohol consumption. Krischler and Glock (2015) studied the effects of different alcohol-related warning frames (i.e., statement warnings, question warnings or control group) on outcome expectations and intention towards drinking less alcohol. The results showed that the statement warnings did not affect intention and outcome expectations. Question warnings however resulted in an increased negative outcome expectations perception but showed no effect on intention. Upon inspecting their stimulus material however, it was found that they made use of closed-ended questions in their question condition as opposed to open-closed-ended questions, which were used in the studies that did find positive effects of self-persuasion (Glock et al., 2013; Müller et al., 2016). Examining the effect of the type of question used within this study and linking these to effects found regarding implicit and explicit conclusions could help explain why Krischler and Glock (2015) found no effects of self-persuasion on intention.
In line with the idea that arguments people generate internally are more
effective than arguments that come from an external source (Mussweiler & Neumann, 2000), it has been found within the field of marketing communication that implicit messages that leave room for interpretation are more effective than explicitly
formulated messages in which the receiver is told exactly what to do (Kardes, Kim & Lim, 1994). Based on this reasoning it can be argued that closed-ended questions
(providing a limited amount of answering options such as yes or no or good or bad) can be associated with an explicit character, whereas open-ended questions (not providing any answer options and thus leaving room for interpretation) indicate a stronger implicit character (Schwarz & Oyserman, 2001).
Within the field of health messages, positive effects of implicit over explicit conclusions can be observed with regard to psychological reactance. Grandpre et al. (2003) found evidence that implicit anti-smoking messages result in less reactance than explicitly stated messages and that respondents evaluated the message and source as more positive in the implicit condition as opposed to the explicit condition. These findings are explained by the fact that people prefer to have options and simply do not like having limited choices (Grandpre et al., 2003). These same results were found in a study towards the effect of implicit and explicit messages regarding marijuana abuse: Adolescents who were exposed to implicit messages rated these as more positive and less controlling than the explicit messages (Burgoon, Alvaro, Broneck, et al., 2002). Furthermore, the explicit messages resulted in a higher intention to try marijuana in the future (Burgoon et al., 2002).
Based on the findings by Burgoon et al. (2002) and Grandpre et al. (2003) regarding implicit and explicit conclusions and the mixed results from the study by Krischler and Glock (2015), it can be expected that a message that activates the process of self-persuasion by using open-ended questions will, due to their implicit character, results in less psychological reactance compared to closed-ended questions. Within the current study, this expectation will be tested by for the first time
combining both closed-ended questions and open-ended questions within one study. This leads to the second hypothesis.
H2: A health warning that is formulated as an open-ended question will lead
to less defensive reactance towards this message compared to a health warning formulated as a closed-ended question, which results in more persuasive effects of the message
Moderating effect of involvement and need for cognition
With regard to the effects of inducing self-persuasion, it can be argued that level of involvement is an important factor in determining whether or not psychological reactance is triggered. According to Brehm and Brehm (1981) reactance increases as the importance of a threatened freedom increases. Within my study this would mean that people who are high involved with the topic of alcohol consumption, are more likely to start elaborating on the topic of responsible alcohol consumption after reading a health warning and are therefore more likely to experience reactance than people who are low involved (Petty & Cacioppo, 1986; Petty, Cacioppo & Schumann, 1983). Furthermore, with regard to type of question, Sawyer and Howard (1991) found that involvement could function as a moderator in the effectiveness of open versus closed-ended advertisements to the extent that high involved consumers respond more favourably to open-ended advertisements compared to low involved consumers. It is therefore expected that involvement functions as a moderator of the effect of type of question on psychological reactance.
H3: The effect of a health warning formulated as a question compared to a
statement on defensive reactance towards this message is moderated by involvement, indicating that this effect is stronger for people who are high involved as opposed to people who are low involved with the topic of alcohol consumption.
H4: The effect of a health warning formulated as an open-question compared
to a closed-ended question on defensive reactance towards this message is moderated by involvement, indicating that this effect is stronger for people who are high involved as opposed to people who have are low involved with the topic of alcohol consumption.
Another potential moderator is need for cognition (NfC), which, within the ELM, can be seen as a predictor of higher elaboration (Cacioppo & Petty, 1982). NfC can be classified as an individual character trait: People with a high NfC experience greater pleasure in thinking and are more likely to engage in elaborate processing, whereas individuals with a low NfC do not like thinking and reasoning and therefore put in less effort to deeply process a message (Cacioppo & Petty, 1982). With regard to the effects of implicit and explicit messages, several studies have looked into the
moderating effect of need for cognition on explicit and implicit conclusions and found that people with a high NfC have more favourable brand attitudes and intention to buy when exposed to implicit conclusions over explicit conclusions (Martin, Lang & Wong, 2003) and more favourable attitudes towards the message as opposed to people with a low NfC (Kao, 2007). Furthermore, the concept of self-persuasion assumes that people will elaborate more deeply on the message (Mussweiler & Neuman, 2000), but this effect is expected to be stronger for people with a high NfC as opposed to people with a low NfC. When people do not elaborate on the message, reactance will not be triggered and therefore it is expected that NfC moderates the effect of a question over a statement and of an open-ended question over a closed-ended question on
H5: The effect of a health warning formulated as a question compared to a
statement on defensive reactance towards this message is moderated by NfC, indicating that this effect is stronger for people who have a higher NfC as opposed to people who have a low NfC.
H6: The effect of a health warning formulated as an open-ended question
compared to a closed-ended question on defensive reactance towards this message is moderated by NfC, indicating that this effect is stronger for people who have a higher NfC as opposed to people who have a low need for
cognition.
Persuasive effects of the health message
The purpose of my study is to increase the persuasive effects of health campaigns by reducing psychological reactance towards these campaigns. So far, it can be
established that the lower psychological reactance that is to be expected by activating self-persuasion should lead to more persuasive effects of the message (Dillard & Shen, 2005). The final dependent variables I took into account to measure these persuasive effects were based on previous studies that have found effects of self-persuasion with regards to other health behaviours. As discussed, previous research regarding smoking cessation has found effects of self-persuasion on intention and risk perceptions (Glock et al., 2013; Krischler & Glock, 2015; Müller et al., 2016), which classifies these variables to also be influenced by the lower psychological reactance caused by self-persuasion within my study towards alcohol consumption.
Furthermore, attitude towards responsible drinking is included in the conceptual model of this study because attitude towards alcohol is a predictor of intention to
consume alcohol based on research regarding the theory of planned behaviour (TPB) (Ajzen, 1988; Huchting, Lac & LaBrie, 2008; Norman & Conner, 2006). Therefore self-persuasion is expected to influence not only intention to drink responsibly, but also attitude towards responsible drinking. Finally, based on the positive message evaluation found by Burgoon et al. (2002) and Grandpre et al. (2003) it is expected that lower defensive reactance caused by self-persuasion results in a more positive attitude towards the health message. The final hypothesis constructed is based on these persuasive effects of the message regarding responsible alcohol consumption.
H7: Lower defensive reactance towards a health warning results in a higher
intention to drink responsibly (a), higher risk perception (b) more positive attitude towards responsible drinking (c) and a more positive attitude towards the message (d).
Method
Design and participants
To test the hypotheses, an online experiment was conducted. This experiment concerned a one-factor (statements vs. closed-ended question vs. open-ended question) between-subjects design. Participants of this experiment were randomly allocated to the statement condition, the closed-ended question condition or the open-ended question condition. Participants were recruited via e-mail and social media and among the participants who completed the experiment a gift card of 25 euro was randomly allocated. In addition, the online university participant recruitment system of the University of Amsterdam was used to recruit participants. In exchange for their contribution, participants received 0.25 participant credits. Because the age from which it is allowed to drink alcoholic beverages in the Netherlands is set to 18, only adults from the age of 18 and older were recruited to participate in this study. The participants who did not complete the online survey were excluded for further analysis (N = 130). The final sample used in this study consisted of a total of 319 participants of which 250 participants were female. The age of the participants ranged from 18 to 75 years old (M = 32.03; SD = 12.14). On average, the participants had consumed 0.88 (SD = 1.15) glasses of alcohol a day in the week before the
experiment, of which men consumed on average more glasses of alcohol a day (M = 1.58; SD = 1.48) than women (M = .69; SD = .96).
Procedure
The participants were recruited through an online invitation to participate in a study on alcohol consumption, which included a hyperlink that lead the participant to the online survey. Before starting the experiment, the participants had to read and sign an
informed consent sheet, after which they were randomly assigned to one of the three conditions. Before being exposed to the manipulation, participants had to answer several demographical questions regarding gender, age and educational level. In addition, the participants answered several questions concerning their NfC and involvement with alcohol. Subsequently, they were allocated to either the statement condition, the closed-ended question condition or the open-ended question condition where they were exposed to a health warning. The participants were told to read the health warning carefully and click the next button when finished reading, a timer was set to five seconds to ensure the participants spent some time observing the poster. Afterwards, the participants filled in a questionnaire with questions concerning their psychological reactance, intention to drink responsibly, risk perceptions, attitude towards responsible alcohol consumption and attitude towards the message. Lastly, several control questions regarding previous alcohol usage were asked. The
questionnaire was identical in all conditions. Finally, the participants were thanked for their participation.
Stimuli
The stimuli used for this study were fictional and were created for the purpose of this experiment. All participants had to read a health poster that was created for a
responsible drinking campaign. To manipulate the level of self-persuasion three types of health posters were created (see Appendix A). The health warning on the poster in the statement condition had the following title: “Excessive alcohol consumption can cause severe health problems!” The health warning in the closed-ended question condition had the title: “Do you think that excessive alcohol consumption is good or bad for your health?” and finally, the open-ended question condition had the title: “What could be the effects of excessive alcohol consumption for your health?” Below
all three versions of the health warning, the guidelines for responsible drinking were incorporated, which was taken from the Guideline for a healthy diet by the Health Council of the Netherlands (2015). The additional information given below the warning stated that men and women are advised to drink no more than one standard glass of alcohol a day. The look, feel and content of the information in the health warning was kept the same in all three conditions, so that possible changes could only be contributed to the manipulation of self-persuasion.
Measures
Psychological reactance. Recent studies in this field demonstrated that
reactance consists of cognitive components and affective components (Dillard & Shen, 2005). The cognitive components measure psychological processes that might result from a persuasive message such as critical processing or counter arguing whereas the affective components measure the negative affect that is experienced by the receiver. Nine items were used to assess both these components of reactance. To assess the cognitive components five items were used from the scale by Dillard and Shen (2005), for example: “The message threatened my freedom to choose” and “The message tried to force an opinion on me”. A 7-point scale will be used to answer these questions ranging from 1 (strongly disagree) to 7 (strongly agree). The affective component was measured by using four items that had been validated in previous studies (Dillard & Peck, 2000; Dillard & Shen, 2005), for example: “While reading the message I felt angry” and “While reading the message I was irritated”. Again, a 7-point scale was used to answer these questions with answer categories ranging from 1 (strongly disagree) to 7 (strongly agree). The Cronbach’s alpha of the cognitive reactance scale was good (x = .88) and the reliability of the affective reactance scale showed to be excellent (x = .90).
Need for cognition. NfC was measured using the 8-item NfC scale by Pieters,
Verplanken and Modde (1987). Items from this scale included statements such as: “I like to be responsible for a situation that requires a lot of thinking” “I would rather do something that requires little thought than something that is sure to challenge my thinking abilities” and “I really enjoy a task that requires thinking of new solutions for problems”. A 7-point scale was used to indicate the extent to which someone agreed with these statements ranging from 1 (strongly disagree) to 7 (strongly agree). Cronbach’s alpha was acceptable (x = .76) and did not increase by deleting items.
Involvement. The involvement with the topic of alcohol was measured using a
7-point scale that consisted of ten semantic differential items taken from the reduced personal involvement scale defined by Zaichkowsky (1985). Participants had to complete the following sentence: “I think that drinking a maximum of one standard glass of alcohol a day is…” The adjectives used to complete this scale were, for example, unimportant-important, means nothing to me-means a lot to me, irrelevant-relevant. Cronbach’s alpha was excellent (x = .91).
Intention to drink responsibly. Intention to drink responsibly was measured
by three 7-point Likert scales that included items based on recommendations by Ajzen (2011). The three items were as follows: “I intend to drink a maximum of one standard glass of alcohol a day in the next two weeks,” “I am sure I will drink a maximum of one standard glass of alcohol a day in the next two weeks,” and “I plan to drink a maximum of one standard glass of alcohol a day in the next two weeks.” A higher score on these scales indicated a higher intention to drink responsibly.
Cronbach’s alpha showed to be excellent (x = .90).
Risk perception. The individual risk someone perceived from drinking more
from Witte et al. (1996). Items on this scale were variations of the following
statement: “If I drink more than 1 standard glass of alcohol a day, I will be at risk for developing health problems.” A higher score on this scale indicated a higher risk perception. Cronbach’s alpha was excellent (x = .90).
Attitude towards responsible drinking. To measure attitude towards
responsible drinking, measures based on Ajzen (2006) were used. A 7-point scale was used, consisting of five semantic differential items which had to be completed by the participants, for example: “If I drink a maximum of one standard glass of alcohol a day, then I think this is … “. The adjectives that had to be used for completing the scales were bad-good, unhealthy-healthy, unwise-wise, unenjoyable-enjoyable, and unpleasant-pleasant. Cronbach’s alpha proved to be excellent (x = .91).
Attitude towards the message. Attitude towards the health poster was
measured using a 10-item semantic differential scale based on an existing scale by Batra and Ahtola (1990), for example “I think the health poster that I just read is…”. The adjectives that had to be used for completing the scales were, for example, unpleasant-pleasant, boring-exciting, not for me-for me. Cronbach’s alpha was good (x = .80).
Control variables. Several control variables were included to control for
possible confounds. These consisted of four questions regarding previous alcohol consumption: “How often did you drink alcohol in the past twelve months?” “How often did you drink five or more glasses of alcohol in the last thirty days?” “How often did you drink two or more glasses of alcohol in the last thirty days?” and “How many glasses of alcohol did you drink each day in the past seven days?” A final question asked for possible reasons participants might have had not to drink anything in the past month, such as medical or religious reasons.
Analysis of procedures
Because this study proposes a moderated mediation model, the Hayes’ (2013) PROCESS macro will be used for the hypotheses testing. First, to test the mediation effect, PROCESS macro in combination with model 4 will be used. This model allows estimating a simple mediation. In addition, the model generates 95% bootstrap confidence intervals (BCI) for the direct and indirect effects and 5,000 bootstrap samples will be used within this estimation method to estimate the bias corrected confidence intervals. To test H3 through H6, model 7 will be used, which allows to test moderation and mediation in one model. For the purpose of this method, two dummy variables will be constructed, the first dummy variable represents the statement condition (0 = statement, 1 = question) and the second represents the type of question condition (0 = closed-ended question, 1 = open-ended question).
Results
RandomizationIn order to determine whether the random assignment to the three experimental groups was successful, several analyses were conducted for the control variables age, gender, educational level, previous alcohol usage and whether someone was a student or not. The results showed that the experimental groups did not differ with respect to age F(2, 315) = 0.95, p = .39, gender χ2 (2) = 1.25, p = .54, educational level χ2 (12) = 13.64, p = .32, previous alcohol usage χ2 (12) = 10.72, p = .55 and the amount of students χ2 (2) = 4.22, p = .12. Based on these results it can be concluded that randomization has been successful.
Descriptive analysis
In order to provide an overview of the dependent variables within this study, a descriptive analysis was conducted of which the results are presented below (see Table 1). The outcome showed that respondents scored around the midpoint for most variables. It is however surprising that the amount of reactance experienced is below the midpoint for both cognitive (M = 3.47, SD = 1.34) as well as for affective
reactance (M = 2.72, SD = 1.35), with affective reactance even being the outcome variable with the lowest mean score. Looking at the outcome variables, most were negatively correlated with cognitive reactance and affective reactance, especially attitude towards the message; respectively (r = -.33) and (r = -.39), and intention to drink responsibly (r = -.16) and (r = -.20).
Table 1
Model correlation matrix with mean and standard deviations
Note: Correlations, means and standard deviations of outcome variables, all variables on
7-point scales. Variables 1 2 3 4 5 6 7 8 1. Cognitive reactance 1.00 - - - - 2. Affective reactance .40 1.00 - - - - 3. NfC -.06 -.02 1.00 - - - - - 4. Involvement .16 .10 .10 1.00 - - - - 5. Intention -.16 -.20 .06 -.14 1.00 - - - 6. Risk perception -.20 -.10 .05 -.12 .13 1.00 - - 7. Attitude alcohol consumption .08 -.11 .10 .32 .21 -.15 1.00 - 8. Attitude message -.33 -.39 .02 .03 .14 .31 .13 1.00 Mean 3.47 2.72 4.87 4.01 3.28 4.69 4.15 3.83
Direct effects
Before testing the model, a MANOVA was conducted to test whether any direct effects could already be observed for a health warning formulated as a statement, a closed-ended question or an open-ended question on any of the outcome variables (see Table 2). The analysis showed that there were no significant direct effects of a health warning formulated as a statement on the outcome variables and neither were there any significant direct effects for a health warning formulated as a closed-ended question or an open-ended question. Being exposed to a question did not result in less defensive reactance as opposed to a statement and neither was there an effect found on any of the outcome variables.
Table 2
Effects of statement, closed-ended question and open-ended question on outcome variables
Hypotheses testing
Effect of question versus statement
With regard to H1, it was expected that a health warning formulated as a question resulted in less defensive reactance, which in turn resulted in more persuasive effects
Statement (N = 103) Closed-ended question (N = 109) Open-ended question (N = 107) M SD M SD M SD F p Cognitive reactance 3.45 1.38 3.51 1.39 3.44 1.25 .09 .91 Affective reactance 2.59 1.30 2.76 1.31 2.82 1.44 .83 .44 Intention 3.27 1.99 3.09 1.82 3.48 2.10 1.05 .35 Risk perception 4.55 1.48 4.78 1.32 4.74 1.33 .84 .43 Attitude 3.97 1.49 4.12 1.55 4.35 1.64 1.55 .21
of the message. Because no direct effects of a question were found on defensive reactance, it was already assumed that defensive reactance did not mediate the effect of a question over a statement on the persuasive effects of the message. This
assumption was confirmed by the mediation analysis. The PROCESS macro was run four times and the results showed that a question did not have a significant effect on cognitive reactance (b = .03, p = .85, BCI [-.29, .35]) and did not significantly affect affective reactance (b = .20, p = .21, BCI [-.11, .52]). Participants exposed to a health warning formulated as a question did not show less defensive reactance compared to participants who were exposed to a health warning with a statement. Because no significant effect of a question on defensive reactance was found, there were also no indirect effects of a question on the persuasive effects of the message: Intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message (see Appendix B, Table B1).
However, with regard to H7, where it was expected that lower defensive reactance resulted in more persuasive effects of the health poster, significant effects were found for the direct effect of cognitive reactance on risk perception, attitude towards responsible drinking, and attitude towards the message (see Table 3A) as well as for affective reactance on intention to drink responsibly, attitude towards
responsible drinking and attitude towards the message (see Table 3B). Based on these findings, H7 was supported: Both lower cognitive and affective defensive reactance resulted in higher intentions to drink responsibly, higher alcohol-related risk
perception and a more positive attitude towards responsible drinking as well as the message.
Table 3A
Direct effect of cognitive reactance in the statement condition on intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message
Variables 95% BCCI
Cognitive reactance
Intention to drink responsibly Risk perception
Attitude towards responsible drinking Attitude towards the message
b SE LLCI ULCI p
-.13 .09 -.31 .04 .13 -.19 .06 -.31 -.07 .00 -.18 .07 .04 .32 .01 -.13 .03 -.20 -.06 .00
Note: BCCI is the bias corrected confidence interval; LLCI is the lower limit of the
confidence interval; ULCI is the upper limit of the confidence interval.
Note 2: A total of 5,000 bootstrap samples were used.
Table 3B
Direct effect of affective reactance in the statement condition on intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message
Variables 95% BCCI
Affective reactance
Intention to drink responsibly Risk perception
Attitude towards responsible drinking Attitude towards the message
b SE LLCI ULCI p
-.24 .09 -.41 -.06 .01 -.03 .06 -.14 .09 .65 -.21 .07 -.35 -.07 .00 -.19 .03 -.26 -.12 .00
Note: BCCI is the bias corrected confidence interval; LLCI is the lower limit of the
confidence interval; ULCI is the upper limit of the confidence interval.
Note 2: A total of 5,000 bootstrap samples were used.
Effect of type of question
In order to test H2, in which it is expected that an open-ended question results in less defensive reactance compared to a closed-ended question, the PROCESS macro with model 4 was run four times. The results showed that type of question did not have a
significant effect on cognitive reactance (b = -.07, p = .69, 95% BCI [-.43, .48]) and did not significantly affect affective reactance (b = .05, p = .77, 95% BCI [-.31, .42]). Participants exposed to a health poster with an open-ended question did not show less defensive reactance compared to participants who were exposed to a closed-ended question. Due to the insignificant effect of type of question on defensive reactance, no indirect effect of type of question on intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message was found (see Appendix B, Table B2). These results suggest that the effect of type of question was not mediated by defensive reactance. Because, according to these findings, an open-ended question did not result in more persuasive effects via defensive reactance than a closed-ended question, H2 was not supported. However, again with regard to H7, direct effects were found for defensive reactance on intention to drink responsibly, attitude towards responsible drinking and attitude towards the message. No significant effects were found for both cognitive and affective reactance on risk perception (see Table 4A & 4B). Based on these results H7 was partially supported.
Table 4A
Direct effect of cognitive reactance in the question condition on intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message
Variables 95% BCCI
Cognitive reactance
Intention to drink responsibly Risk perception
Attitude towards responsible drinking Attitude towards the message
b SE LLCI ULCI p
-.12 .11 -.33 .39 .28 -.14 .07 -.28 .01 .07 .18 .09 .01 .35 .04 -.14 .04 -.22 -.05 .00
Note: BCCI is the bias corrected confidence interval; LLCI is the lower limit of the
confidence interval; ULCI is the upper limit of the confidence interval.
Table 4B
Direct effect of affective reactance in the question condition on intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message
Variables 95% BCCI
Affective reactance
Intention to drink responsibly Risk perception
Attitude towards responsible drinking Attitude towards the message
b SE LLCI ULCI p
-.28 .10 -.48 -.08 .01 -.09 .07 -.23 .05 .19 -.24 .08 -.41 -.08 .00 -.02 .10 .81 .23 .00
Note: BCCI is the bias corrected confidence interval; LLCI is the lower limit of the
confidence interval; ULCI is the upper limit of the confidence interval.
Note 2: A total of 5,000 bootstrap samples were used.
Moderating effect of involvement
To test H3 and H4, where the mediated effect of a question over a statement and an open-ended question over a closed-ended question was expected to be moderated by the level of involvement with alcohol consumption, model 7 of the PROCESS macro was used (Hayes, 2013). The results showed that the effect of a question instead of a statement as well as the type of question on both cognitive and affective reactance was not moderated by involvement (see Table 5).
For both H3 and H4, it was also tested whether the level of involvement with alcohol moderated the indirect effect of a question over a statement and the type of question on the dependent variables. The results showed that the indirect effect of a question versus a statement and the type of question on intention to drink responsibly, risk perception, attitude towards responsible drinking and towards the message, through cognitive and affective reactance was not moderated by involvement (see Appendix B, table B3 & B4).
Table 5
Effect of question versus statement, and type of question on defensive reactance (cognitive and affective) moderated by involvement
Note: BCCI is the bias corrected confidence interval; LLCI is the lower limit of the
confidence interval; ULCI is the upper limit of the confidence interval.
Note 2: A total of 5,000 bootstrap samples were used.
Both the results for H3 and H4 indicated that the effect of a question versus a statement and the effect of type of question (open-ended versus closed-ended) on defensive reactance did not differ for respondents with higher levels of involvement compared to respondents with lower levels of involvement and neither did the indirect effect of a question versus a statement and type of question on the dependent
variables. Based on these results, H3 and H4 were not supported.
Need for cognition
In order to answer H5 and H6, in which NfC was expected to moderate the effect of a question over a statement and the type of question (open-ended vs. closed-ended) on defensive reactance, the PROCESS macro with model 7 was used again. The results showed that the effect of a question versus a statement as well as the type of question on both cognitive and affective reactance was not moderated by NfC (see Table 6).
Variables 95% BCCI
Interaction effect of involvement
Question vs. statement -> cognitive reactance Question vs. statement -> affective reactance Type of question -> cognitive reactance Type of question -> affective reactance
b SE LLCI ULCI p
-.12 .13 -.38 .15 .38 .04 .13 -.23 .30 .78 .11 .15 -.19 .41 .47 .27 .16 -.04 .57 .09
Table 6
Effect of statement versus question, and type of question on defensive reactance (cognitive and affective) moderated by NfC
Variables 95% BCCI
Interaction effect of NfC
Question vs. statement -> cognitive reactance Question vs. statement -> affective reactance Type of question -> cognitive reactance Type of question-> affective reactance
b SE LLCI ULCI P -.02 .19 -.40 .36 .90 -.29 .27 .10 .19 .23 .24 -.67 -.19 -.37 .09 .14 .72 .23 .57 .17
Note: BCCI is the bias corrected confidence interval; LLCI is the lower limit of the
confidence interval; ULCI is the upper limit of the confidence interval.
Note 2: A total of 5,000 bootstrap samples were used.
Because no moderation effect was found, there was also no moderation of NfC on the indirect effect of the statement condition and type of question on any of the outcome variables (see Appendix B, Table B5 & B6). These findings indicated that the effect of a question versus a statement and the type of question on defensive reactance did not differ for respondent with higher NfC compared to respondents with lower NfC and neither did the indirect effect on the outcome variables. Based on these results, H5 and H6 were not supported.
Additional analyses
Some unexpected results were found in the current study, one of them being that involvement does not influence the effect of a question over a statement or the effect of type of question (open-ended versus closed-ended) on defensive reactance.
However, suggesting that involvement does not play a role in the activation of defensive reactance is contradicting with theory. Because of this, additional analyses
were conducted to see whether involvement influenced defensive reactance and the dependent variables directly. The results show that involvement had a significant direct effect on cognitive reactance (b = .18, p < .05, 95% BCI [.06, .30]) but not on affective reactance (b = .11, p = .07, 95% BCI [-.01, .24]). Participants with higher levels of involvement with alcohol experienced significant more defensive cognitive reactance compared to participants with lower levels of involvement with alcohol. Because significant effects were found of involvement on one of the components of defensive reactance, it was also tested whether involvement had an indirect effect on the dependent variables through defensive reactance. The results showed that
involvement had a significant indirect effect via defensive reactance on intention to drink responsibly, risk perception and attitude towards the message. No significant indirect effect was found for involvement on attitude towards responsible drinking (see Table 7).
Table 7
Indirect effects of involvement on intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message (via defensive reactance)
Note: BCCI is the bias corrected confidence interval; LLCI is the lower limit of the
confidence interval; ULCI is the upper limit of the confidence interval.
Note 2: A total of 5,000 bootstrap samples were used.
Variables 95% BCCI
Involvement -> def. reactance -> intention Involvement -> def. reactance -> risk perception Involvement -> def. reactance -> attitude alcohol Involvement -> def. reactance -> attitude message
b SE LLCI ULCI -.05 .03 -.11 -.00 -.03 -.00 -.05 .02 .02 .02 -.09 -.04 -.09 -.01 .04 -.01
Based on these findings it was concluded that higher levels of involvement result in more defensive reactance, which in turn results in lower intentions to drink
responsibly, lower risk perceptions and a more negative attitude towards the message.
Conclusion and Discussion
Although several studies within the field of health communication have conducted research with regard to the benefits of formulating health warnings into questions instead of statements based on the theory of self-persuasion, none of these studies have included a measure of defensive reactance within their model (Glock et al., 2013; Krischler & Glock, 2015; Müller et al., 2009, 2016). In addition, no study within this field of research has included both closed-ended questions as well as open-ended questions within one experiment.
The research aim of this study was threefold. First, its purpose was to research the mediating effect of defensive reactance, triggered by self-persuasion, on the persuasive effects of the message (intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message). Second, this study tried to extent current research on self-persuasion by incorporating both closed-ended as well as open-closed-ended questions within this study. The third aim of this study was to investigate whether the level of involvement with alcohol and the level of NfC influence the effect of a question over a statement or the type of question (closed-ended versus open-(closed-ended) on defensive reactance and the persuasive effects of the message.
With regard to the first aim, the results from the conducted experiment show that, in contrast with previous research (Glock et al., 2013; Müller et al., 2009, 2016) and the expectations within this study, being exposed to a question instead of a
statement did not affect defensive reactance, intention to drink responsibly, risk perception, attitude towards responsible drinking and attitude towards the message. No mediating effect of defensive reactance was found. In addition, with regard to the second aim, type of question (either closed-ended or open-ended) also did not have an effect on both defensive reactance and the persuasive effects of the message. As with the statement condition, no mediating effect of defensive reactance was found for the effect of type of question on the persuasive effects of the message. However,
significant effects were found for the direct effect of defensive reactance on the outcome variables, indicating that lower defensive reactance does result in more persuasive effects of the message (Dillard & Shen, 2005). With regard to the final aim, the results demonstrate that involvement does not moderate the effect of self-persuasion and type of question on defensive reactance and neither does NfC. In contrast, involvement does have a positive effect on the persuasive effects of the message through defensive reactance: The higher the involvement with alcohol, the more defensive reactance is experienced, which results in lower intentions to drink responsibly, lower risk perceptions and a more negative attitude towards the message.
These contradicting findings can possibly be explained by the fact that previous studies only found an effect for health warnings formulated as a question with regard to smoking cessation (Glock et al., 2013; Müller et al., 2009, 2016) whilst limited effects were found for alcohol consumption (Krischler & Glock, 2015). The health poster used within this study targeted responsible alcohol consumption, which may feel less threatening than an anti-smoking warning for smokers, due to the fact that drinking alcohol is perceived as more acceptable behaviour than smoking.
Because people did not feel threatened by the message in the first place, formulating a health warning as a question instead of a statement did not influence the effectiveness
of the health poster via defensive reactance (Brehm & Brehm, 1981; Dillard & Shen, 2005).
An alternative explanation for the absence of an effect of self-persuasion and the type of question can be found in the fact that in contrast to previous studies (Glock et al., 2013; Müller et al., 2016) this study included an additional line of text below the question on the health poster participants were exposed to, which informed the participants about the guidelines for responsible alcohol consumption. This additional line of text was not formulated as a question, and could have therefore intervened with the effect that the question should have had. Despite being aware of the fact that this additional text was not included in previous studies, I still decided to include this guideline in my stimulus material. The main reason for this inclusion was to reach a higher level of ecological validity within my study instead of solely conducting a ‘proof of principle’ study.
Even though the possibility of this additional information to influence the effect of the questions was made as small as possible (by making this text smaller as opposed to the questions or statement, and posting this information at the bottom of the poster instead of in a central place), it cannot be guaranteed that this information did not distract attention from the questions or statement the participants were also exposed to. Besides the format of this additional information in the poster, the content could have also played a role in influencing the effect of the questions or statement. Because the Health Council of the Netherlands has only relatively recently
(November, 2015) changed the guidelines for responsible alcohol consumption to a maximum of one standard glass of alcohol a day for both men and women, the information could have come as a surprise to most participants within this study. When participants were not aware of the guidelines before participating in this study,
the information regarding the maximum of one glass of alcohol could have drawn more attention than the questions or statement they were exposed to as well. Because of this, the effect of the manipulation might have lost its power.
This can also explain why a direct effect of involvement was found but no interaction effect of involvement. The expected moderating effect of involvement was probably offset by the lack of effects found for a question over a statement as well as for the type of question. Therefore, when solely analysing the direct effect of
involvement the results do show an effect of involvement, namely that people who are more involved with alcohol experience higher levels of defensive reactance. These results are in line with the ELM (Petty & Cacioppo, 1986; Petty, Cacioppo & Schumann, 1983). High-involved readers process the message more deeply and are therefore more prone to experience reactance as opposed to low-involved readers.
Theoretical and practical implications
Even though these results show some contradiction with previous research, the outcome of this study does contribute to existing work on self-persuasion in several ways. First of all, this study is among the first to test the concept of self-persuasion by using stimulus material with a higher ecological validity as opposed to solely
portraying a question or a statement. The results show that the concept of
self-persuasion does not seem to show persuasive effects when paired with additional text such as guidelines. This is an important theoretical implication, since this finding suggests a specific design of the stimulus material when studying the concept of self-persuasion. Furthermore, practical implications can be given regarding the content of campaign material to health institutions wanting to trigger a process of
Second, this study is the first within this field of research to include both open-ended and closed-open-ended questions within the same experiment. Previous studies have included either open-ended questions (Glock et al., 2013; Müller et al., 2009, 2016) or closed-ended questions (Krischler & Glock, 2015). The results from this current study suggest that open-ended questions are not more effective than closed-ended questions with regard to reducing defensive reactance and increasing persuasive effects of a health warning regarding alcohol consumption. This finding adds to the current literature and suggests that the lack of effects found in the study by Krischler and Glock (2015) cannot be contributed to the fact that their manipulation material consisted of closed-ended questions instead of open-ended questions.
Limitations and suggestions for future research
Although this study gives insight into the mechanisms through which self-persuasion does or, in this case, does not seem to work, my study represents a first initial attempt to test the effects of type of question on health warnings regarding alcohol
consumption and therefore knows some limitations. First, the high percentage of women within this study (i.e., 78.4%) might have influenced the results due to the fact that women tend to drink less alcohol than men (Gezondheidsraad, 2015). This was also found to be true within my study: The results show that women in my sample on average drink less than one standard glass of alcohol a day and thus already conform to the guidelines, while men on average drink more than one glass of alcohol a day. It is therefore advised to replicate the study amongst a sample in which gender is more evenly distributed.
A second limitation of this study can be found with regard to the stimulus material. Although health messages usually induce a sense of fear and can be perceived as threatening (Dillard & Shen, 2005), the health poster used in the
experiment was not pre-tested to determine whether participants indeed experienced the message as threatening. This links back to the explanation for the lack of effects found in the current study due to the non-threatening nature of health warnings for alcohol consumption. A suggestion for future research is therefore to conduct a pre-test to evaluate the extent to which people perceive the message to be threatening. In line with this reasoning, future research should replicate this study for the topic of smoking cessation instead of alcohol consumption to investigate whether effects of open-ended questions over closed-ended questions can indeed be found for a more threatening behaviour. Hereby it is advised to also include a measure of psychological reactance because a mediation effect is still to be expected, especially for a more threatening behaviour such as smoking (Brehm, 1966).
Despite the above-mentioned limitations, the results of this study are of great importance for research regarding the concept of self-persuasion within the field of health communication. The current study provides insight into the mechanisms through which the concept of self-persuasion works. In addition, the results underline the importance of decreasing psychological reactance within health communication in order to obtain more persuasive effects of a threatening message, especially among high-involved individuals.
Acknowledgements
I would like to express my appreciation to my thesis supervisor Prof. Dr. Bas van den Putte for his support throughout this process. Many thanks for assisting me by
References
Ajzen, I. (1988). Attitudes, personality and behavior. Milton Keynes: Open University Press.
Ajzen, I. (2006). Constructing a theory of planned behavior questionnaire. Retrieved from http://www-unix.oit.umass.edu/∼aizen/pdf/ tpb.measurement.pdf
Ajzen, I. (2011). Behavioral interventions: Design and evaluation guided by the theory of planned behavior. In M. M. Mark, S. I. Donaldson & B. Campbell (Eds.), Social psychology and evaluation (pp. 74–100). New York, NY: Guilford Press.
Baldwin, A., Rothman, A., Vander Weg, M., & Christensen, A. (2012). Examining causal components and a mediating process underlying self-generated health arguments for exercise and smoking cessation. Health Psychology, 32, 1209-1217. doi:org/10.1037/a0029937
Batra, R., & Ahtola, O. T. (1990). Measuring the hedonic and utilitarian sources of consumer attitudes. Marketing Letters, 2, 159–170.
Bem, D. J. (1965). An experimental analysis of self-persuasion. Journal of Experimental Social Psychology, 1, 199−218.
Bem, D. J. (1967). Self-perception: An alternative interpretation of cognitive dissonance phenomena. Psychological Review, 74, 183−200.
Blankenship, K. L., & Craig, T. Y. (2006). Rhetorical question use and resistance to persuasion: An attitude strength analysis. Journal of Language and Social Psychology, 25(2), 111–128.
Brehm, J. W. (1966). A theory of psychological reactance. New York: Academic Press.
Brehm, S. S., & Brehm, J. W. (1981). Psychological reactance: A theory of freedom and control. New York: Academic Press.
Burgoon, M., Alvaro, E. M., Broneck, K., Miller, C., Grandpre, J. R., Hall, J. R., et al. (2002). Using inter-active media tools to test substance abuse prevention messages. In W. D. Crano & M. Burgoon (Eds.), Mass media and drug prevention: Classic and contemporary theories and research (pp. 67-87). Mahwah, NJ: Lawrence Erlbaum.
Cacioppo, J. T., & Petty, E. R. (1982). The need for cognition. Journal of Personality and Social Psychology, 42, 116–131.
Dillard, J. P., & Peck, E. (2000). Affect and persuasion: Emotional responses to public service announcements. Communication Research, 27, 461–495. Dillard, J. P., & Shen, L. (2005). On the nature of reactance and its role in persuasive
health communication. Communication Monographs, 72, 144 - 168. Glock, S., Müller, B. C., & Ritter, S. M. (2013). Warning labels formulated as
questions positively influence smoking-related risk perception. Journal of Health Psychology, 18, 252–262. doi:10.1177/1359105312439734
Godin, G., Bélanger-Gravel, A., Vézina-Im, L.-A., Amireault, S., & Bilodeau, A. (2012). Question-behaviour effect: A randomised controlled trial of asking intention in the interrogative or declarative form. Psychology & Health, 27, 1086–1099. doi:10.1080/08870446.2012.671617
Grandpre, J., Alvaro, E. M., Burgoon, M., Miller, C. H., & Hall, J. R. (2003).
Adolescent reactance and anti-smoking campaigns: A theoretical approach. Health Communication, 15, 349−366.
https://www.gezondheidsraad.nl/sites/default/files/201524_richtlijnen_goede_ voeding_2015.pdf
Hanewinkel, R., Sargent, J. D., Poelen, E. A., Scholte, R., Florek, E., Sweeting, H., . . . Morgenstern, M. (2012). Alcohol consumption in movies and adolescent binge drinking in 6 European countries. Pediatrics, 129, 709–20. doi:10.1542/peds.2011-2809
Harris, P. R., & Epton, T. (2009). The impact of self-affirmation on health cognition, health behaviour and other health-related responses: A narrative review. Social and Personality Psychology Compass, 3, 962-978.
Harris, P. R., Mayle, K., Mabbott, L., & Napper, L. (2007). Self-affirmation reduces smokers’ defensiveness to graphic on-pack cigarette warning labels. Health Psychology, 26, 437–446. doi:10.1037/0278-6133.26.4.437
Hayes, A. F. (2013). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. New York, NY: Guilford Press. Hoch, S. J., & Deighton, J. (1989). Managing what consumers learn from experience.
Journal of Marketing, 53, 1-20.
Hornik, R., Jacobsohn, L., Orwin, R., Piesse, A., & Kalton, G. (2008). Effects of the national youth anti-drug campaign on youths. American Journal of Public Health, 98, 2229–2236.
Huchting, K., Lac, A., & LaBrie, J. W. (2008). An application of the theory of
planned behavior to sorority alcohol consumption. Addictive Behavior, 33, 538–551.
Kardes, F. R., Kim, J., & Lim, J. (1994). Moderating effects of prior
knowledge on the perceived diagnosticity of beliefs derived from implicit versus explicit product claims. Journal of Business Research, 29, 219-24.
