The strength of suggestion
—————————————————————
How people’s resistance to forceful language affects brand evaluations and behavioural outcomes as contingent on their customer status
Kim Wassenaar
Master Thesis Persuasive Communication University of Amsterdam
Supervisor: Saar Mollen June 24th, 2016
Abstract
Language forcefulness has been found to influence the effectiveness of health messages. Explicit messages generally lead to more resistance than implicit messages and
subsequently to lower levels of behavioural intentions. We know little however about the exact resistance strategies that people engage in when they are confronted with an explicit message or of how these may affect outcomes beyond behavioural intentions. This study shed light on the resistance strategies that people use in response to implicit and explicit messages and the effects of these strategies on brand expertise, brand trustworthiness, behavioural attitude, and behavioural intention. Moreover, the study looked at how
prospective and current customers may differ in their reactions to language forcefulness. In an online experiment, 239 participants were divided over two two experimental conditions, one containing an implicitly voiced message, and the other an explicit message. After exposure to one of these two messages, their scores on nine distinct resistance strategies, two measures of brand outcomes, and two measures of behavioural outcomes were measured. The experiment showed that explicit language leads to more resistance than implicit language, and consequently to less positive brand and behavioural outcomes, and that these effects are contingent on customer status. Prospective customers were found to react more negatively to explicit language than current customers. The two groups were also shown to vary in terms of the types of resistance strategies they used.
Introduction “Fat kids become fat adults”
“Big bones didn’t make me this way. Big meals did” (Kliff, 2012).
Are these claims, that were part of an obesity awareness campaign in Georgia a number of years ago, offensive or merely truthful? Do they evoke feelings of resistance or is their unequivocal message impossible to ignore? Facts aside, the campaign against childhood obesity featuring the above and more similarly outspoken statements stirred up quite some dust when it was first
launched. Experts from different fields questioned whether the campaign was not doing more harm than good, but Children’s Healthcare of Atlanta (the institution behind the advertisements) found it was time to stop sugarcoating the issue. But was their approach effective? The campaign certainly achieved its goal of raising awareness of the obesity problem in the state (Matzigkeit, 2012), but conclusive statistics on its results were never published and the main sentiment seems to be that the campaign was predominantly believed to be hurtful and offensive (Obesity Action Coalition, 2016).
While message appeals are no new area of research, the specific topic of tone of voice has been relatively understudied. Yet, the handful of studies that have addressed the topic have highlighted its importance by demonstrating its profound effects on outcomes such as
persuasiveness, reactance, attitude and message acceptance (Buller et al., 2000; Burgoon, Jones, & Stuart, 1975; Miller, Lane, Deatrick, Young, & Potts, 2007; Quick & Considine, 2008). Tone of voice has been operationalised in many ways. The use of a conversational human voice has received much attention within the context of organisational communication on social media platforms (Noort & Willemsen, 2012; Park & Cameron, 2014; Park & Lee, 2013; Yang, Kang & Johnson, 2010), and has also been studied in terms of for example lexical concreteness, which refers to the level of specificity of a text (Miller et al., 2007), and language intensity (Ball & Goodboy, 2014; Dillard & Shen, 2005; Miller et al., 2007; Steindl, Jonas, Sittenthaler,
study will focus on the latter operationalisation of tone of voice: language intensity, also known as language forcefulness.
The level of language forcefulness of a text is dependent on two elements: how controlling it is, and how demeaning it is. The higher a text scores on both of these elements, the more forceful it can be considered to be. High levels of language forcefulness have been found to increase the levels of reactance audiences experience and consequently their levels of resistance. Reactance refers to a psychological state (Brehm, 1966), characterised by cognitive and emotional processes such as anger and negative thoughts (Dillard & Shen, 2005; Quick & Stephenson, 2008; Rains & Turner, 2007). To reduce the experienced reactance, audiences engage in resistance, which can manifest itself as a number of different strategies (Brehm, 1966; Miller et al., 2013). Examples of such strategies are source derogation, selective exposure and social validation (Fransen, Verlegh,
Kirmani, & Smit, 2015). Lower levels of language forcefulness, because they induce less reactance and thus elicit less resistance, have been shown to increase the chances of message acceptance as compared to highly forceful messages (Dillard & Shen, 2005).
Research, however, is scarce and inconclusive. While some studies find negative effects of high intensity language messages (Dillard & Shen, 2005) other studies find positive effects (Buller et al., 2000; Miller et al., 2007). An explanation for these opposing findings may be found in the attitudes of the recipients of the message and more specifically in the extent to which these are positive, and in their extant behavioural intentions (Buller et al., 2000; Hong, 2011). Another explanation could relate to the level of importance they ascribe to the behaviour in question (Baek, Yoon, & Kim, 2015; Hong, 2011). When an audience deems an issue of high importance, and holds attitudes towards the promoted behaviour that are in line with those in the message they are exposed to, forceful messages may be more effective than implicit messages.
Practically speaking, this implies that there is not a one size fits all “best practice” when it comes to the tone that is adopted in persuasive messages. This is of importance not only to governmental health campaigns, but also to the increasing number of commercial organisations
such as Personal Body Plan and Rockstar Lifestyle in the Netherlands (Personal Body Plan, 2016; Rockstar Lifestyle, 2016) and Tracy Anderson in America and the United Kingdom (Tracy Anderson, 2016) that these days are invested with improving our health and changing our habits through numerous fitness and nutrition programs. Many of these companies use social networking sites (SNS) to communicate and engage with their audiences. Audiences that are made up of prospective and current customers alike. Current customers of these programs have plausibly already made changes to their lifestyle and as such may hold attitudes that are more aligned with those held by the company behind the program, than do people who are interested in change, but are not quite there yet. Based on earlier findings that currently held attitudes, issue importance and intentions may moderate which tone of voice (explicit versus implicit) is most effective (Baek et al., 2015; Buller, Borland, & Burgoon, 1998; Buller et al., 2000; Kronrod, Winstein, & Wathieu,
2012a), it seems conceivable that these customers in different stages will perceive messages differently.
The contradictory findings concerning the effectiveness of forceful language and the scarce research on attitude and intention as possible moderators of and explanations for these paradoxical findings leads to conclude that more insight is needed into the processes that underlie these effects. Additionally, as more and more commercial companies involve themselves with health
communication, metrics beyond persuasiveness become of interest as for these companies marketing outcomes may be of equal if not higher importance. Thus far, most studies on the
interplay between language forcefulness, resistance and behaviour have been done in the context of non-commercial health communication messages (Baek et al., 2015; Buller et al., 2000; Dillard & Shen, 2005; Jenkins & Dragojevic, 2011; Mourre & Gurviez, 2015; Quick & Bates, 2010; Quick & Considine, 2008; Quick, Scott, & Ledbetter, 2011). Although Quick and Kim (2009) did study the effects of forceful language in advertising, they did not test any commercial outcomes, but merely looked at the effects of forceful language on reactance and resistance. This study aims to contribute to this existing body of literature by carrying the topic of language forcefulness and resistance into
the commercial realm, by including outcome measures more important to for-profit companies, specifically brand evaluations in the form of brand expertise and brand trustworthiness.
Additionally, it will test the potentially moderating effects of currently held attitudes, issue
importance, and intentions as mentioned above by capturing them in one variable, namely people’s customer status, in an applied context. Lastly, by measuring resistance not as a single construct but as separate resistance strategies, this experiment gives insights into which resistance strategies relate to which brand and behavioural outcomes, and how different types of customers may differ in the strategies they put to use. To further increase the practical relevance of this study, the
experiments will be conducted for one specific company: Personal Body Plan (PBP). The company is currently one of the main players on the Dutch market in its field and offers its customers a six month plan to health based on four central pillars: behaviour, nutrition, training and recovery. Through their various social media outlets and weblog they reach over 40,000 people daily,
providing them with information that could help their audience become a healthier, better version of itself.
All in all the current study will address how different levels of language forcefulness affects brand evaluations and behavioural outcomes as a result of the resistance strategies they elicit, as well as the hypothesised moderating role of customer status on these effects. The following research question captures this study’s aim:
RQ: How does language forcefulness affect brand evaluations and behavioural outcomes through invoked resistance, and in how far are these effects subject to one’s relationship to the message’s source?
Theoretical Background
This study draws on psychological reactance theory (Brehm, 1966) to explain the way in which tone of voice is expected to influence brand attitudes and behavioural intentions and the role of resistance strategies and customer status in this process. I will first review the extant literature on
resistance strategies and how these affect brand and behavioural outcomes. Then the concept of language forcefulness and its relationship to resistance will be discussed. Lastly, I will explore existing literature as well as politeness theory (Brown & Levinson, 1978), and cognitive dissonance theory (Festinger, 1957) to hypothesise how current and prospective customers may differ in their reactions to forceful language.
Resistance
In his psychological reactance theory (PRT), Brehm (1966) states that any persuasive message has a chance of and will most likely elicit some level of reactance (Brehm, 1966). PRT posits that people hold certain freedoms which they are driven to protect and maintain. These freedoms can pertain to many things, such as actions, emotions, and attitudes (Brehm, 1966; Wicklund, 1974). However, they must always be conscious freedoms, in as such that a person must be aware of possessing a freedom for it to qualify as such (Dillard & Shen, 2005), because if one is not aware of being in possession of any given freedom, one cannot feel threatened in the execution of its freedoms. This perceived threat is a necessary condition for reactance, as reactance occurs when an individual experiences a threat to one of his or her freedoms. Differently put, people value their freedom to be able to at any time choose between various options (Quick & Bates, 2010). When they feel they are being limited or influenced in executing this freedom, they will experience reactance.
The fourth central concept in reactance theory, freedom, threat to freedom, and reactance being the first three, is restoration of freedom. It refers to the attempts that people make to claim back whichever freedoms they feel are being threatened and is aimed at giving them back a sense of autonomy and self-determination (Miller, Lane, Deatrick, Young & Potts, 2007). Restoration of freedom can be categorised into direct restoration (i.e. the performance of the behaviour one feels is being threatened), indirect restoration (i.e. evaluating the threatened behaviour more positively), and vicarious restoration (the performance of the threatened behaviour by a person similar to oneself) (Brehm, 1966; Dillard, & Shen, 2005; Miller et al., 2007; Miller et al., 2013).
In its core, restoration of freedom is a form of resistance. Resistance refers to an
unwillingness to change, or a want to maintain the status quo (Knowles & Linn, 2004). It is dual in nature, in as such that it can refer both to an outcome (i.e., not conforming to influences of change), and a motivational state (i.e., the disposition to resist any pressure to change). This distinction relies partly on methodical grounds, where resistance as a motivational state assesses the process of resistance, whereas resistance as an outcome evaluates only the results. This distinction is important to make, as occurrence of resistance as a motivational state does not guarantee the presence of resistance as an outcome. In other words: just because someone is motivated to resist a message, does not mean that the message’s result in terms of for example attitude change or behavioural intention is indeed altered. This research includes resistance as both a motivational state (i.e. the resistance strategies people engage in) and an outcome (i.e. the brand and behavioural outcome measures). Doing so allows us to assess not only the effects of language forcefulness on both operationalisations of resistance, but also the impact of resistance as a motivational state on
resistance as an outcome, by assessing the effectiveness of different resistance strategies in enabling actual resistance as a result.
Resistance Strategies. In their attempts to resist any persuasive message, people can engage in various resistance strategies. These can take a number of forms and can broadly be divided into three categories: avoiding, contesting, and empowering strategies (Fransen, Verlegh, Kirmani, & Smit., 2015). Avoidance strategies involve attempts to not be exposed to a particular message. They can be either physical (e.g. leaving the room to not have to hear a conversation), mechanical (e.g. zapping to a different television channel), or cognitive (e.g. skimming over arguments that are not in line with one’s views, also known as selective attention). People engage in contesting strategies when they challenge a message in terms of either its content, source, or persuasive tactics. The contesting category concerns attempts to contend and repudiate either the message itself, or the source behind the message. Examples of contesting strategies are counter arguing (i.e. assessing the arguments in a message, and coming up with counter arguments to refute those), and source
derogation (i.e. questioning the source behind a message in terms of their expertise or
trustworthiness, thereby dismissing their credibility) (Fransen, Smit, & Verlegh, 2015) Lastly, empowering involves attempts to reassure oneself of one’s existing attitude. This can be achieved through efforts such as attitude bolstering (i.e. consciously coming up with arguments that support the reader’s currently held attitude), social validation (i.e. thinking of significant others who share one’s own beliefs and herein finding validation for one’s own behaviour and attitude), and self-assertion (i.e. reminding oneself of one’s confidence in the currently held beliefs, and the fact that these should not and cannot possibly change) (Jacks & Cameron, 2003). Looking at individual resistance strategies rather than at resistance as a whole, provides opportunities that go beyond the scope of most studies to date. It gives insight into the exact strategies people employ in response to a message as well as the effectiveness of each individual strategy by measuring its impact on brand and behavioural outcomes, thereby assessing the impact of resistance strategies (i.e. resistance as a motivational state) on actual resistance (i.e. resistance as an outcome).
So far, we have mentioned two factors that cause people to engage in resistance strategies: threats to freedom (i.e. experienced reactance), and a dispositional unwillingness to change, or in other words a desire to maintain the status quo. Fransen, Smit and Verlegh (2015) identify a third antecedent to resistance which they call concerns of deception. This factor builds on the notion that people dislike being wrong. They want to belief that the attitudes and beliefs they hold are correct, and do not like to be misled. The extent to which this factor can be found to cause resistance is, amongst other things, dependent on the level of persuasion knowledge held by an individual, previous experiences with persuasion attempts, skepticism, and lastly message characteristics. Each of the three antecedents to resistance as identified by Fransen, Smit & Verlegh (2015) - threats to freedom, unwillingness to change, and concerns of deception - has been found to correlate with specific resistance strategies as they were categorised above. Threats to freedom were found to lead to contesting and empowerment strategies, concerns of deception to lead to contesting strategies, and reluctance to change to lead to empowerment and biased processing strategies, and all three
antecedents result in avoidance strategies. By in hindsight connecting the resistance strategies different people engage in, we should on these grounds thus be able to speculate about which antecedents are most relevant for either type of customer and level of language forcefulness. For example, as will be further elaborated on below, it is likely that current customers will show less unwillingness to change than do prospective customers due to their current attitudes and
involvement. The discussion section of this paper will discuss this in more depth.
Language Forcefulness
One of the many factors that has been found to influence increase threats to freedom and thus the level of resistance people engage in, is the tone that is adopted in a persuasive message (Miller et al., 2007). The focus of the current research specifically will be on tone in terms of the forcefulness of a piece of writing. A forceful tone of voice, also known as high-intensity language (Buller et al., 2000; Burgoon, Jones, & Stewart, 1975), controlling language (Miller et al., 2007), or an explicit tone (Grandpre, Alvaro, Burgoon, Miller, & Hall, 2003; Miller, Burgoon, Grandpre, & Alvaro, 2006) is seen to convey a single meaning and leaves little doubt as to the source’s intentions. Highly forceful language is characterised by the use of imperatives rather than propositions, and of forceful adverbs such as “ought”, “must”, and “should” (Miller et al., 2007; Miller et al., 2013). Forceful language can be perceived as demeaning (Jenkins & Dragojvic, 2011), and a text written in this tone will very clearly, directly, and briefly tell a person what to do. It entails little ambiguity. On the other end of the spectrum is language that is little forceful, also known as implicit language, or low-intensity language. Implicit language is highly ambiguous and much more open to
interpretation than its explicit counterpart. Because messages written in this tone make more use of qualifiers such as “perhaps”, “possibly”, and “maybe”, and emphasise self-initiation and choice, they tend to come across as politer. In the remainder of this paper, we will discuss ‘language forcefulness’ as the general indicator of the level of forcefulness in a message. To differentiate between the two levels of language forcefulness that will be tested in this experiment, ’implicit’ will
be used to refer to low levels of language forcefulness and ‘explicit’ will indicate high levels of language forcefulness.
Language forcefulness has been studied in the context of numerous health topics. Quick and Considine (2008) manipulated four messages encouraging people to start exercising either
individually or as a group activity, so that two messages contained an explicit tone and two messages contained an implicit tone. The titles of their four manipulations clearly show the distinction between the two levels of language forcefulness, with the explicit messages containing the phrase ‘you have to do it’, whereas the implicit messages started with the word ‘consider’. Jenkins and Dragojevic (2011) chose an entirely different topic as the subject of their messages (i.e. dental flossing), but their manipulations show similarities with those of Quick and Considine (2008) and our current research. Their implicitly voiced message contained phrases such as ‘…you might want to think about making flossing a regular habit.’, whereas in the explicit message this same line was phrased saying ‘other severe problems […] make[s] it just stupid not to floss every single day of your life.’.
Forceful Language and Resistance
Forceful language has repeatedly been found to lead to higher levels of perceived threats to freedom than does implicit language (Miller et al. 2007; Quick & Considine, 2008), and thus to higher levels of resistance. Moreover, Buller et al. (2000) found implicit language led to higher levels of
improvement in the promoted behaviour (i.e., sun protection), than explicit language did. Jenkins and Dragojevic (2011) conducted two experiments and concluded that less forceful messages produced higher levels of behavioural intention towards the promoted behaviour than did highly forceful messages.
An explanation for these findings can be found in politeness theory (Jenkins & Dragojevic, 2011). Politeness theory posits that people have two types of wants: negative face wants, which are similar to people’s need for autonomy and refer to our desire to be free of limitations and
obligations in the choices we make, and positive face wants, which refer to our desire to be
positively evaluated by others, to be liked, and to be seen as competent (Brown & Levinson, 1978). A message is seen as polite when it supports both of these wants (Dillard et al., 1997; Holtgraves, 1997; Lim & Bowers, 1991).
Because forceful language limits people’s choices and as such increases their perceived threat to freedom, and is moreover considered to be highly face threatening in terms of both negative face (in as such as that it is limiting) and positive face (as it can be seen to question one’s authority) (Jenkins & Dragojevic, 2011), it is likely to be experienced as impolite. Because polite messages are more likely to be accepted, this is one explanation for the finding that explicit messages lead to more resistance than implicit messages.
At the same time however it seems conceivable that explicit language, which makes relatively clear the communicator’s intentions, would be considered as less deceptive than implicit language, and as such would be received more positively and thus is more likely to be accepted (Miller et al., 2007). However, the dominating trend in the findings of previous studies implies that the controlling nature of explicit messages seems to override this positive effect.
While the effects of reactance and resistance as a whole have been frequently studied, not much is known about the individual resistance strategies in which people engage as depending on the level of forcefulness of the message they are confronted with. Not much more is known about the consequent effects on brand and behavioural outcomes. Where resistance in general has been found to negatively influence such outcomes, we can so far say little about which exact strategies are most effective in instilling resistance (as an outcome) to persuasive messages. The current research will explore these topics, by examining the effects of language forcefulness on brand and behavioural outcomes per specified resistance strategy.
H1a: The use of explicit language (versus implicit language) will lead to lower evaluations of brand expertise than the use of implicit language because it leads to higher levels of resistance.
H1b: The use of explicit language (versus implicit language) will lead to lower evaluations of brand expertise than the use of implicit language because it leads to higher levels of resistance.
H1c: The use of explicit language (versus implicit language) will lead to lower behavioural attitudes than the use of implicit language because it leads to higher levels of resistance. H1d: The use of explicit language (versus implicit language) will lead to lower behavioural intentions than the use of implicit language because it leads to higher levels of resistance.
Differences in Customer Status
There is a great number of factors that influence level of resistance people show in response to a message. Persuasion knowledge, previous experiences with persuasion attempts, and skepticism are just a few examples (Fransen, Smit, & Verlegh, 2015; Friestad & Wright, 1994; Koslow, 2000; Van Reijmersdal, Rozendaal, & Buijzen, 2012). Other factors that have been found to correlate with resistance are issue importance, source credibility, attitude of acceptance (i.e. the distance between the position towards a given topic as it is being communicated in a message, and the position held by the reader of this message towards the discussed topic). Current and prospective customers of PBP may be expected to vary on these factors, which could influence the way in which they react to implicit versus explicit messages. As such, prospective and current customers can be hypothesised to differ in the levels of resistance and strategies they engage in when presented with differently voiced messages. The three aforementioned factors underlying this expectation will be more elaborately discussed below.
Issue importance. For reactance and thus resistance to occur, the threatened behaviour must be relevant to the individual (Chandler, 1990). After all, if the freedom under attack is not valued or used, there will be no perception of threat. Following this line of reasoning, it has been found that people who are highly invested with a topic react differently to explicit messages than do people who have lower levels of involvement. More specifically, forcefully voiced messages are more
effective when people deem the discussed topic to be of importance as compared to implicit messages, while the reverse holds for people do not consider a topic to be important (Baek et al., 2015; Buller et al., 2000; Kronrod et al., 2012).
Source credibility. Research based on the language expectance theory has shown that high credibility sources have a greater linguistic freedom than do low credibility sources (Burgoon, Denning & Roberts, 2002): they can adopt a wider range of communicative strategies without running the risk of compromising on persuasiveness. People who are engaged in an ongoing
relationship with a source such as PBP have put a significant part of their lives into the hands of this organisation. In the case of this specific organisation, to be a customer means to have entered into a six month contract with the company as the cost of €80,- a month. During these six months,
customers are assigned a personal coach with whom they are in touch on a daily basis. They receive guidance in establishing healthy eating and exercise patterns, but also discuss highly personal matter. As such, it is highly plausible that they have more confidence and trust in this organisation than people who are not yet engaged on such a high level. On that account, current customers are likely to react more positively to highly explicit messages than prospective customers who have arguably not yet reached the same levels of trust and perceived credibility with regards to the company. This latter group imaginably reacts more positively to implicit messages, due to their presumed lower evaluations of brand credibility.
Cognitive dissonance theory and currently held beliefs. A final process that leads to the expectation that current and prospective PBP members will differ in their reactions to explicit and implicit messages stems from cognitive dissonance theory (CDT). CDT posits that people can experience cognitive inconsistencies (or dissonance), which result in psychological discomfort (Festinger, 1957). These inconsistencies refer to discrepancies between the beliefs and/or behaviour and beliefs. When these inconsistencies are sufficiently substantial (i.e. when the inconsistency is important enough, and out of balance as compared to the number of consistencies one holds), a person will be motivated to resolve the inconsistency. This can happen in three ways: by changing
elements of a message to make it more consonant (e.g. selective attention), by adjusting the ratio of consonant to dissonant beliefs, or by changing the level of importance that is given to the topic under consideration (O’Keefe, 1990). Generally, people will seek to reduce the experienced dissonance in the easiest way possible (Cameron, 2008). This often means they will actively avoid messages that do not conform with their current attitudes and beliefs, in other words: they engage in resistance strategies such as selective exposure or selective attention (D’Alessio & Allen, 2002).
To illustrate, Buller and colleagues (1998) found that people who already had a readily formed intention to engage in sun protective behaviours - in other words: they believed engagement in sun protective behaviours to be positive and important - responded more positively toward forcefully voiced messages than people without these intentions.
Proposing a conditional mediation model
As mentioned, language forcefulness increases perceived threats to freedom and fuels people to engage in resistance strategies. These resistance strategies in turn have been found to increase resistance as an outcome. In this study, this out come will be measured in terms of brand
evaluations (brand expertise and brand trustworthiness), and behavioural outcomes (behavioural attitude and behavioural intention). I also assessed extant research on three factors that have been found to influence the way in which people react to explicit messages: issue importance, source credibility, and currently held beliefs as hypothesised on the basis of cognitive dissonance theory. These theories and earlier research findings lead to the expectation that explicit language will have a more negative effect on brand and behavioural outcomes than implicit language because it causes higher levels of resistance. Also, I expect there will be a difference between prospective and current customers in the resistance strategies they employ and consequently the extent to which their brand attitudes and behavioural intentions are influenced by the level of language forcefulness in a message. Hypotheses 1a and 1b capture the prediction that the the effect of language forcefulness on brand and behavioural outcomes will be contingent on the status of the customer that is exposed
to the message, because these two types of customers differ in the extent to which they engage in resistance in response to an explicit or implicit message.
H2a: For prospective customers, the use of explicit language (versus implicit language) has a stronger negative effect on evaluations of brand expertise and Brand trustworthiness than for current customers as a result of a stronger increase in resistance for prospective
customers than for current customers.
H2b: For prospective customers, the use of explicit language (versus implicit language) has a stronger negative effect on behavioural attitudes and behavioural intentions than for current customers as a result of a stronger increase in resistance for prospective customers than for current customers.
Whilst a general prediction of the effects of language forcefulness on brand and behavioural outcomes through resistance can be made, as well as the expectation that there will be differences between current and prospective customers in this regard, a lack of current knowledge does not allow for more specific predictions of how these two types of customers may differ in their use of specific resistance strategies, or of the effectiveness of these individual strategies. This experiment will shed light on these questions by answering the following explorative question:
RQ1: How do current and prospective customers differ in terms of the resistance strategies they engage in?
Taken together, the above hypotheses and research questions compose a model
hypothesising a two-step relationship between language forcefulness and brand and behavioural outcomes via resistance. Moreover, these effects are suspected to be affected by whether people arecurrent customers of the company communicating the either implicitly or explicitly voiced message. The full conditional mediation model is depicted in Figure 1.
Method Design and Sample
This study uses a combined-strategy experiment with a 2 (Language Forcefulness: implicit versus explicit) x 2 (Customer Status: prospective customer versus current customer) factorial design to test the aforementioned research questions and hypotheses. Every participant was exposed to either an implicit or an explicit piece of writing and later categorised as either a current or a prospective participant of PBP. The latter variable was thus not manipulated but measured. The text and questions used in the experiment were written in Dutch to prevent miscommunication and issues with the comprehension of the message and questions. Hence, the material that can be found in Appendix D is in Dutch.
A total of 11.436 people was approached by email and asked to participate in the current experiment. The email was sent to people who are registered to the PBP newsletter and are thus either prospective or current customers of the company. In the email, an incentive (a chance to win a DoorMeal food package at the value of €28,-) for participation was given. The email included a link that when clicked, took people to the online survey which, as was indicated in the email, took
Language Forcefulness Resistance Brand Outcomes Customer Status Behavioural Outcomes
around ten minutes to complete. The survey was set-up using Qualtrics software, which is available online. The first page of the online experiment included the ethical terms as set by the Amsterdam School of Communication Research to which all participants had to agree before being able to start the experiment. The ethical disclosures as they were included in the experiment can be found in Appendix B. The response rate was 3.33%. A total of 381 people started the experiment and 250 of these (65.62%) fully completed it.
Of the 250 participants that completed the experiment, two people were excluded for being under the age of eighteen, and one person was excluded for having entered ’99’ as their age. An additional eight people were excluded because they indicated that they had previously been a customer of PBP, but were no longer, and itwas unclear to which category they should belong in terms of their involvement with and commitment to the company. This resulted in a final sample of 239 participants, 78.20% of which was female (n = 187) and 21.80% of which male (n = 52), ranging in age from 18 to 62, with an average age of 29.82 (SD = 8.43). Of these participants, 139 people were current customers of PBP and 101 people were prospective customers. Randomisation of the participants over the two experimental conditions was successful. There were no significant differences between the participants in the two conditions in terms of their gender (F(1, 238) = 3.56, p = .060), age (F(1, 238) = .39, p = .536), educational level (F(1, 238) = .22, p = .637). In the final sample, 120 participants were exposed to the implicit condition and 119 participants were exposed to the explicit condition.
Procedure
The first page of the experiment in Qualtrics included a brief introduction to the experiment, again stating the estimated time people would need to complete the study, as well as what to expect: a short text of around 150 words, a number of questions pertaining to this text, and several questions concerning PBP. Participants were also assured that their participation would be anonymous and their information treated confidentially. This first page also included the ethical disclosure forms as
mentioned above and included in Appendix A. Once people had signed to agree to participation in the experiment, they could continue to the second page. This page merely mentioned that they would next be shown a brief text, and asked participants to please read this text carefully. The next page then showed one of the two manipulated texts as shown in Appendix B. A ten second timer was included on this page to prevent people from (accidentally) forwarding to the next page before having thoroughly read the text. The pages following the stimulus material included the
measurement instruments as presented in Appendix D. Items measuring resistance were presented first, followed by items measuring the four outcome variables (brand expertise, brand
trustworthiness, behavioural attitude, and behavioural intention), demographic questions and lastly items measuring involvement and customer status. In the questionnaire, several additional measures were included that were chosen to further analyse in the experiment due to practical constraints. Finally, people were asked what they thought the aim of the experiment was and were given a chance to enter their email address if they wanted to have a chance of wining the DoorMeal package. The last page included a short debriefing, informing participants of the aim of the experiment and inviting them to contact me in case of any questions or concerns.
Material
For this experiment, two conditions were created by manipulating a short text so that it was written in either an implicit or explicit tone. The text was drafted from scratch, taking earlier messages from PBP as a guide in terms of length and topic. The messages in both conditions were about how people can prepare themselves for a day out and adhere to their healthy lifestyle through the use of ‘meal prepping’, which refers to the habit of preparing and bringing your own meals when knowing you will not be home to eat. This topic was chosen as it is gender-neutral and the habit achievable to implement for everyone regardless of age or lifestyle. Moreover, from my personal experience as a coach for PBP I know that many current participants have not yet made a habit of meal prepping, meaning they are as much subject to change as are prospective customers.
The texts in the two conditions differed only in terms of their levels of language
forcefulness. The font size, displayed pictures, and choice of words other than those changed to facilitate the manipulation of language forcefulness were consistent across the two conditions. The level of language forcefulness was manipulated on the basis of examples and operationalisations of this concept in prior research (Jenkins & Dragojevic, 2011; Jones, Burgoon & Stewart, 1975; Miller et al., 2007). For the implicit condition, qualifiers such as ‘perhaps’, ‘maybe’, and ‘possibly’, and propositions, such as ‘could’, ‘might’, and ‘may’, were frequently used. Through these word choices, emphasis was laid on choice and self-initiation. The resulting message was rather polite and ambiguous, not making very clear what the sources intentions are (i.e. is it merely meant informatively, or does the message try to persuade?). The explicitly voiced message, in contrast, was characterised by the use of imperatives and forceful adverbs such as ‘should’ and ‘must’. The intent of this message is clear (i.e. trying to persuade people to adopt the habit of meal prepping) and very clearly tells people what to do. To illustrate, consider the following example of a line that was used in both texts, manipulated according to the above mentioned guidelines.
Implicit: You could try to bring your own meals when you know you are going to be away from home for a day.
Explicit: [To avoid being dependent on the offer on site,] you must bring your own meals when you are going to be away from home for a day.
The messages were shown as they would appear on the Facebook page of PBP (see
Appendix B). Facebook was chosen as it is one of PBP’s most prominent communication channels and allows for longer pieces of written information than does Instagram, the company’s other main social outlet.
Pre-test. Prior to the experiment, a pre-test was conducted. People were approached through personal Facebook channels and asked whether they would participate in an under five-minute survey. When clicked on the distributed link, they were redirected to the experiment and given a brief introduction of what to expect. This introduction told them they would be reading a short text
and subsequently asked eight questions related to this text. As participants were approached through Facebook the response rate is unknown, but all 73 people who started the questionnaire, completed it. Of these 73 people, 38.40% were male (n = 28) and 61.60% were female (n = 45), ranging in age from 18 to 38 (M = 24.92, SD = 4.53). The pre-test included four manipulations of language
forcefulness: highly implicit, moderately implicit, moderately explicit, and highly explicit. The manipulations of these texts followed the guidelines written out above. The four manipulated texts that were used in the pre-test can be found in Appendix C. Every participant was shown one of these four texts. The texts were not integrated into an image as were the final two manipulations used in the actual experiment, but were shown as plain text in Qualtrics.
The pre-test included seven 6-point bipolar questions measuring the forcefulness of the language in the four messages: 1) forceful-not forceful; 2) commanding-not commanding; 3) brief-ambiguous; 4) clear-unclear; 5) refutable-irrefutable; 6) direct-indirect; 7) polite-impolite (see also Appendix E). As no instrument exists to measure perceptions of language forcefulness, these items were drafted on the basis of the main differences between implicit and explicit language as
described in extant literature and as used as guidelines for the manipulation of the stimulus material. A factor analysis showed the seven items to load onto two factors. The first factor (clarity)
consisted of items three through six, had an eigenvalue of 2.40, explained 34.31% of the variance, and showed a reliability of α = .76. The second factor (intensity), consisting of items one, two and seven, had an eigenvalue of 2.28, explained 32.60% of the variance, and showed a reliability of α = .80. An eighth item was included to measure the credibility of the message. This item was measured on the same 6-point bipolar scale as the preceding items, and ranged from ‘credible’ to ‘incredible’.
ANOVA analyses showed that there was no significant effect of language forcefulness on the clarity scale (F(3, 69) = 1.09, p = .357, η = .05), but a significant effect was found of language forcefulness on the intensity scale (F(3, 69) = 8.831, p < .000, η = .28). Participants in different conditions thus differed in their evaluations of the forcefulness of the read message, when
showed that participants in the highly explicit condition rated the message as most forceful (M = 3.86, SD = 1.08) and participants in the highly implicit condition rated their read message as least forceful (M = 5.30, SD = 1.07). Bonferonni post-hoc tests showed this difference to be significant (p < .000) The two moderate conditions also differed, with people in the moderately explicit condition scoring higher on experienced forcefulness (M = 4.24, SD = 0.86) than people in the moderately implicit condition (M = 5.27, SD = 1.22). The differences between the two moderate conditions were also found to be significant (p = .037).
Because the differences in mean scores between the two implicit conditions and the moderately explicit condition were only small, the choice was made to use the highly explicit condition in the final experiment. With the differences between the moderately implicit and highly implicit condition being non-significant (p = 1.000), for the implicit condition the moderate
manipulation was used. This choice was made on the basis that the highly implicit condition was evaluated as somewhat less credible (M = 2.61, SD = .98) than the moderately implicit condition (M = 2.31, SD = 1.40), albeit not significantly (p = 1.000).
Proposed moderator
As mentioned before, the second independent variable ‘customer status’, our proposed moderator, was not manipulated but only measured. At the end of the questionnaire, participants answered a question inquiring whether they were currently registered as a participant of PBP. They were given three possible answers: ‘Yes’, ‘No’, and ‘No, but I have been in the past’ and were categorised according to their answers. Customer status was used as a dichotomous variable in all PROCESS analyses, with 0 = ‘current customer’ and 1 = ‘prospective customer’. The expectation that current and prospective customers would differ in their reactions to implicit and explicit messages was in part based on the presumption that they would vary in terms of their involvement with health and the level of importance they assign to this topic. This was however not found to be te case. The two groups of customers did not show any differences in terms of the importance they assigned to health
(F(1, 239) = .69, p = .407) or their level of involvement with this subject (F(1,239) = .217, p = .642) . Nevertheless, because of the importance of customer status for the model, and the possibility that the two groups would be distinct on other grounds, the choice was made to continue with the analyses as planned.
Measures
Resistance.To measure the resistance strategies participants engaged in, Fransen, Ter Hoeven & Verlegh’s (2013) 26 item measurement construct, containing two to four questions per resistance strategy, measuring ten strategies total, was used. For this research, one resistance strategy was excluded (avoidance) because the set-up of this experiment did not allow participants to engage in this strategy. Also, the item ‘I remind myself of the fact that this advertisement is trying to sell me something’ that was originally the fourth item measuring ‘invoking persuasion knowledge’ was left out because it was not applicable to the current experiment. Furthermore, minor changes were made to the phrasing of the original questions to ensure a fit with the current research, which does not contain advertisements - as was assumed by the original construct -, but informative texts. For example, the second item measuring the resistance strategy ‘attitude
bolstering’ was changed from ‘I think about facts that support my own opinion about the advertised product’ to ‘I thought about facts that support my own opinion about the discussed behaviour’. The statements were translated from English to Dutch and further alterations were made to the
formulation of the three questions measuring message derogation to reduce the level of
suggestiveness in these questions. One of the original questions was phrased stating ‘I think about how misleading the ad is’, and was changed to ‘I thought about whether I consider this message to be misleading’. Changes were also made to the formulation and structure of the questions
measuring the last resistance strategy, distraction, to ensure a better fit with the current set-up. For example, the first item was changed from ‘I think about things that are unrelated to the ad’ to ‘I thought about things that weren’t related to the topic discussed in the message’. The ensuing
measurement instrument can be found in Appendix D. Factor analyses showed that all items per individual strategy measured one construct (there was only one eigenvalue higher than one). The results of the factor and reliability analyses for the nine separate resistance strategies that were measured in this study can be found in Appendix E.
Brand evaluation. To measure people’s attitudes towards PBP after having been exposed to either the implicit or explicit text, Jack and Cameron’s (2003) ‘brand evaluation’ measurement scales, which divide the construct into ‘brand expertise’ and ‘brand trustworthiness, were used. Both scales were measured using 7-point bipolar scales. Brand expertise consisted of eight items, including comparisons such as ‘unwise - wise’, ‘incompetent - competent’, and
‘uninformed/informed’. All eight items were found to indeed measure one construct with an
eigenvalue of 5.46, explaining 68.30% of the variance. The scale proved highly reliable (α = .93, M = 5.73, SD = .92). Brand trustworthiness was measured by six items, including
‘unfriendly/friendly’, ‘bad/good’, and ‘cold/warm’, and showed a eigenvalue of 4.53, explaining 75.46% of the variance. This scale, too, proved highly reliable (α = .93, M = 5.66, SD = 1.08). The full measurement construct can be found in Appendix D.
Behavioural attitude. To assess participants’ attitude towards the behaviour promoted in the message (i.e., meal prepping), five items were included in the questionnaire. These items were formulated on the basis of the theory of planned behaviour as originally suggested by Ajzen (2013) and measured on 7-point bipolar scales. All questions were introduced with ‘If in the coming months I were to prepare my meals when going out for a day, this would be..’ and included
contrasts such as ‘bad/good’, and ‘negative/positive’. The five items loaded onto one factor with an eigenvalue of 4.01, explaining 80.24% of the variance. The resulting scale was found to be highly reliable (α = .94, M = 6.03, SD = .97).
Behavioural intention. To measure the intention of people to actually perform the
promoted behaviour, and thus the chances of people engaging in meal prep on days on which they are not at home at meal time, three items were included. These items were set to measure
behavioural intention as proposed by Ajzen (2013). The items were again formulated according his guidelines and resulted in the following three statements:
1. In the coming months, I will try to prepare my meals when I am going out for a day. 2. I plan to prepare my meals when I am going out for a day in the coming months. 3. I the coming months, I will prepare my meals when I am going out for a day.
The items were measured on 7-point Likert scales ranging from ‘Unlikely’ to ‘Likely’, or in the case of the second item for language purposes from ‘Most certainly not’ to ‘Certainly’. The three items measured one construct with an eigenvalue of 2.72, which explained 92.40% of thevariance. This scale, too, proved highly reliable (α = .96, M = 5.19, SD = 1.48).
Health involvement. Lastly, in order to be able to test whether current participants of PBP indeed differed from prospective customers in their level of involvement with health topics, the three items, all measured on 7-point Likert scales ranging from ‘Very unimportant’ to ‘Very important’ were added to the questionnaire. All three questions were introduced with ‘How
important do you think it is to..’, followed by ‘to eat healthily’, ‘to exercise frequently’, and ‘to live healthily’. The three questions were found to indeed load onto one factor (EV = 2.47, IVC =
82.33%) and make for a highly reliable scale (α = .89).
Results
To test the validity of the model proposed in Figure 2, 36 moderated mediation analyses were run using Hayes’ (2012) PROCESS macro, selecting model 7. The results of these analyses will be discussed below per outcome variable and strategy. A full report of all the relationships analysed with the PROCESS macro by Hayes (2012) can be found in tables 1 through 4 The pathways a, b, c, and d that are described in these tables are depicted in Figure 2. Pathway a describes the direct effect of the independent variable, language forcefulness (two levels: implicit versus explicit), on our hypothesised mediator, resistance, which is subdivided into the nine individual resistance strategies. Pathway b describes the direct effect of one resistance strategy on one of the four
outcome variables (i.e. brand expertise, brand trustworthiness, behavioural attitude, and behavioural intention). Pathway c’ is the direct effect of language forcefulness on one of the four outcome variables when controlled for customer status (two levels: prospective (PC) versus current (CC) customers) and one out of nine resistance strategies. Pathway d depicts the interaction between customer status and language forcefulness in relation to the latter’s effect on resistance. It shows the differences in resistance between prospective and current customers. Because both language
forcefulness (i.e. implicit versus explicit) and customer status (i.e. prospective versus current) are dichotomous, the coefficient of d indicates the difference in the effect of explicit (versus implicit) language on resistance for current as compared to prospective customers. Lastly, the proposed mediating result of resistance for the effect of language forcefulness on brand expertise, brand trustworthiness, behavioural attitude, and behavioural intention will be separately given for PC and CC. Table 1 provides the results for all nine models (one per resistance strategy) including brand expertise as a dependent variable. Apart from the coefficients and significance of pathways a, b, c’, and d, the table includes the coefficients of the indirect effect of language forcefulness on brand expertise per group of customers (i.e. prospective or current) per resistance strategy. Behind ‘model’, the overall significance of the full conditional mediation model for one of the nine
resistance strategies and with brand expertise as dependent variable can be found. Tables 2, 3, and 4 show these same effects for the remaining three dependent variables: brand trustworthiness,
behavioural attitude, and behavioural intention, respectively.
The effect of language forcefulness (implicit versus explicit) on the nine resistance
strategies (pathway a) and the interaction effect between customer type (prospective customers (PC) versus current customer (CC)) and language forcefulness (pathway d) on each of the nine resistance strategies are constant for all four outcome variables. In other words, these values are the same for the model testing the conditional effect of language forcefulness on brand expertise through resistance, as they are for the models testing the conditional effects of language forcefulness through resistance on brand trustworthiness, behavioural attitude, and behavioural intention.
Therefore, their exact values will be mentioned in text only in the first section which discusses the models on brand expertise. In later sections, I refer you to tables 1 through 4 for these values. To enhance readability, the tested models will not be fully written out, but will be abbreviated. Each model will be referred to by a combination of the proposed mediator and the tested outcome
variable. To exemplify, ‘the CS-IPK-BE’ model will refer to the model hypothesising that customer type moderates the effect of language forcefulness (LF) on brand expertise (BE) via invoking persuasion knowledge (IPK) as a resistance strategy. See Figure 3 for a full overview of the models and respective abbreviations.
Conditional Indirect Effects of Language Forcefulness on Brand Expertise (BE)
Invoking persuasion knowledge (IPK). The full CS-IPK-BE model, suggesting an effect of language forcefulness on brand expertise through IPK as moderated by customer status, did not proof significant (CI = [-.00, .23]). Pathway c’ was found to be significant (c’ = -.32, p = .013), meaning there was a direct effect of language forcefulness on brand expertise in such a way that explicit messages led to lower BE (M = 5.54, SD = 1.00) than implicit messages (M = 5.90, SD = .79) when controlled for variations in IPK. Pathway a was significant: language forcefulness was found to positively affect IPK (a = 1.71, p < .000), as explicit messages invoked more use of IPK (M = 4.44, SD = 1.30) than implicit messages (M = 3.26, SD = 1.48). The use of IPK as a resistance strategy was however not found to significantly affect evaluations of brand expertise (b = -.08, p = .074), meaning pathway b was insignificant. The interaction effect (pathway d) was significant (d = -.94, p = .011), meaning current and prospective customers differed in their use of IPK in reaction to explicit language. The increase in IPK as a result of exposure to an explicit (versus implicit) message was less strong for current customers than for prospective customers. IPK significantly mediated the negative effect of language forcefulness on brand expertise for current customers (CC = .06, CI = [.16, .00]). For prospective customers, this mediated effect was not significant (PC = -.13, CI = [-.32, .01]). This means that while forceful language led to a stronger activation of
persuasion knowledge and consequently a decline of brand expertise evaluations for current
customers, this was not the case for prospective customers. The model suggesting that the effects of language forcefulness on brand expertise are dependent on one’s customer status and a result of the use of IPK is however not overall a reliable predictor of the variation in scores on brand expertise.
Counter arguing (CA). The overall CS-CA-BE model proved significant (CI = [.18, .66]), meaning that variations in brand expertise are significantly predicted by differences in the use of counter arguing as a result of language forcefulness and customer status. There was no
significant effect of language forcefulness on BE when controlled for variations in CA (c’ = -.27, p = .170). Explicit language had a significant positive effect on the use of CA (a = 1.79, p < .000). People exposed to the explicit message showed higher levels of CA (M = 3.61, SD = 1.47) than participants in the implicit condition (M = 2.67, SD = 1.21). Pathway d was also found to be
significant in a negative direction (d = -.14, p < .000), meaning prospective customers scored higher on resistance than current customers when exposed to an explicit message. Higher CA was found to lead to a significant decrease in brand expertise (b = -.27, p < .000). For prospective customers, the use of counter arguing significantly decreased their evaluations of brand expertise when they were exposed to the explicit message as compared to the implicit message (PC = -.49, CI = [-.74, -.30]). For current customers however, the use of counter argumentation had no significant mediating effect (CC = -.11, CI = [-.24, .01]). Counter argumentation thus significantly explains the negative effects of explicit language on brand evaluations for prospective customers, but not for current customers. The overall significance of the model means that the interplay between customer status and counter arguing explains a greater proportion of the variety in scores on brand expertise as a result of language forcefulness than does a model without these predictors.
Attitude bolstering (AB). The CS-AB-BE model did not prove significant (CI = [-.12, .10]), and thus customer status and AB together do not significantly better predict variations in brand expertise than does a model without these predictors. Pathway a did prove significant (a = 1.42, p < .000). Explicit language leads to an increase in AB (M = 5.17, SD = 1.16) as compared to
F ig ur e 3. A n o v er v ie w o f th e 3 6 m o d el s te st ed w it h H ay es ’ (2 0 1 2 ) P R O C E S S m ac ro L an gu ag e F o rc ef ul ne ss (L F ) (i m p lic it ve rs us e xp lic it) In vo k in g P er su as io n K no w le d ge (I P K ) C o un te r A rg ui n g (C A ) A tt itu d e B o ls te ri n g (B A ) S el ec tiv e E xp o su re ( S E ) S o ur ce D er o ga tio n (S D ) S o ci al V al id at io n (S V ) A ss er tio ns o f C o nf id en ce (A o C ) M es sa ge D er o ga tio n (M D ) D is tr ac tio n (D ) O u tc o m e V ar ia b le s a) B ra nd E xp er ti se b ) B ra nd T ru st w or th in es s c) B eh av io ur al A tt it ud e d ) B eh av io ur a l I nt en ti o n C us to m er S ta tu s (p ro sp ec tiv e ve rs us c ur re nt )
Table 1
Model Coefficients for the Conditional Process Model in Figure 2 for Brand Expertise
Resistance strategy
Outcome Measures Indirect effects of X on Y via M per value of W [95% BCBCI]
coeff se t p coeff se LLCI ULCI
Invoking Persuasion Knowledge
a. 1.81 .27 6.59 .000 PC -.11 .08 -.29 .05
b. -.06 .04 -1.41 .161 CC -.05 .04 -.14 .01
c’ -.29 .13 -2.29 .023
d. -1.03 .36 -2.87 .005 index se LLCI ULCI
model .07 .06 -.00 .23
Counter Arguing
a. 1.79 .27 6.61 .000 PC -.49 .11 -.74 -.30
b. -.27 .04 -6.71 .000 CC -.11 .06 -.24 .01
c’ -.16 .12 -1.38 .170
d. -.14 .35 -3.96 .000 index se LLCI ULCI
model .38 .12 .18 .66 Attitude Bolstering
a. 1.42 .28 .500 .000 PC .02 .06 -.10 .13
b. -.01 .04 .24 .809 CC .00 .01 -.01 .04
c’ -.41 .13 -3.32 .001
d. -1.28 .37 -3.49 .001 index se LLCI ULCI
model -.01 .05 -.12 .10
Selective Exposure
a. .92 .29 3.19 .002 PC -.12 .06 -.26 -.04
b. -.14 .04 -3.16 .002 CC -.01 .03 -.07 .07
c’ -.35 .12 2.96 .003
d. -.88 .37 -2.36 .019 index se LLCI ULCI
Table 1
Model Coefficients for the Conditional Process Model in Figure 2 for Brand Expertise
Resistance strategy
Outcome Measures Indirect effects of X on Y via M per value of W [95% BCBCI]
coeff se t p coeff se LLCI ULCI
Source Derogation
a. 1.63 .29 5.60 .000 PC -.61 .15 -.90 -.33
b. -.37 .03 -11.03 .000 CC -.17 .09 -.38 -.01
c’ -.08 .10 -.81 .422
d. -1.17 .38 -3.11 .002 index se LLCI ULCI
model .44 .16 .14 .77 Social Validation
a. 1.25 .31 3.98 .000 PC -.13 .07 -.30 -.03
b. -.11 .04 -2.75 .006 CC -.02 .03 -.10 -.04
c’ -.35 .12 2.93 .004
d. -1.19 .41 -2.71 .007 index se LLCI ULCI
model .12 .07 .02 .31 Assertions of Confidence
a. 3.33 .18 18.24 .008 PC .02 .03 -.04 .10
b. .03 .05 .60 .550 CC .01 .02 -.01 .07
c’ -.43 .12 -3.49 .001
d. -.48 .24 -1.38 .169 index se LLCI ULCI
model -.01 -.03 -.10 .02
Message Derogation
a. 1.60 .31 5.16 .000 PC -.47 .11 -.72 -.27
b. -.29 .04 -8.28 .000 CC -.28 .08 -.46 -.14
c’ -.06 .11 -.51 .614
d. -.65 .50 -1.61 .108 index se LLCI ULCI
Table 1
Model Coefficients for the Conditional Process Model in Figure 2 for Brand Expertise
Resistance strategy
Outcome Measures Indirect effects of X on Y via M per value of W [95% BCBCI]
coeff se t p coeff se LLCI ULCI
Distraction
a. .85 .24 3.52 .001 PC .20 .09 -.41 -.06
b. -.24 .05 -4.81 .000 CC .20 .05 -.12 .08
c’ -.32 .12 -2.77 .006
d. -.77 .31 -2.44 .015 index se LLCI ULCI
model .18 .10 .03 .44 Notes. BCBCI = bias-corrected bootstrap confidence interval using 10.000 bootstrap samples; significant effects are bold; a = the direct effect of X on M; b = the direct effect of M on Y; c’ = the direct effect of X on Y as controlled for W and M; d = the interaction effect of X and W on M; PC = prospective customers; CC = current customers; N = 233.
implicit language (M = 4.40, SD = 1.68). This effect is significantly moderated by customer status (d = -1.28, p = .001) in as such that current customers show a smaller increase in AB than do prospective customers in reaction to exposure to an explicit (versus implicit) message. The use of AB does not affect brand expertise (b = -.01, p = .809), meaning that use of this resistance strategy does not result in resistance as an outcome. No significant mediation effects of AB were found for either type of customers (PC = -.02, CI = [-.10, .13]; CC = .00, CI [-.01, .04]), meaning that for neither current nor prospective customers, the use of attitude bolstering explains the significant negative effects of explicit as compared to implicit language on brand expertise (c’ = -.41, p = .001).
Selective exposure. The full CT-SE-BE model was significant (CI = [.02, .29]). Explicit language was found to lead to higher levels of SE (a = .92, p = .002; M = 3.68, SD = 1.41) than implicit language (M = 3.25, SD = 1.38), and this effect was found to vary dependent on customer status (d = -.88, p = .019). Again, the increase in SE as a consequence of exposure to an explicit versus implicit message was smaller for current customers than for prospective customers. SE had a significant negative effect on evaluations of brand expertise (b = -.14, p = .002). It significantly
mediates the relationship between language forcefulness and brand expertise for prospective customers (PC = -.12, CI = [-.26, -.04]), but not for current customers (CC = -.01, CI = [-.07, .07]). There was a direct negative effect of language forcefulness on brand expertise even when
controlling for SE and customer status (c’ = -.35, p = .03). Overall, SE and customer status are significant predictors of the effects of language forcefulness on brand expertise. SE is used as a form of resistance by prospective customers, but not by current customers.
Source derogation. The full CT-SD-BE model proved significant (CI = .14, .77]).
Together, language forcefulness, source derogation and customer status thus are better predictors of the variation in brand expertise than a model without these predictors. Explicit language was found to lead to significant higher levels of SD (M = 2.56, SD = 1.68) than implicit language (a = 1.63, p < .000; M = 1.71, SD = 1.02). The significant interaction effect between customer status and language forcefulness (d = -1.17, p = .002) indicates that prospective customers showed a greater increase in SD when they were exposed to an explicit as compared to implicit message than did current customers. Current customers’ scores on SD thus increase less when exposed to an explicit message as compared to an implicit message than do prospective customers’ scores. Higher levels of SD were found to lead to significant lower levels of brand expertise (b = -.37, p < .000). Source derogation is responsible for the effect of language forcefulness on brand expertise for both current and prospective customers (CC = -.17, CI = [-.38, =-.01]; PC = -.61, CI = [-.90, -.33]). The
mediated negative effect of explicit as compared to implicit language on brand expertise was thus found to be much stronger or prospective customers than for current customers. A direct effect of language forcefulness on brand expertise was not found when controlling for SD and customer status (c’ = -.08, p = .422).
Social validation (SV). The full CS-SV-BE model proposing SV as a mediator of the effect of language forcefulness and brand expertise and customer status as a moderator of pathway a (i.e. the effect of language forcefulness on social validation) was significant (CI = [.02, .31]). Explicit language led to significantly more use of SV (a = 1.25, p < .000; M = 2.81, SD = 1.25) than implicit
language (M = 2.17, SD = 1.46). This effect is different for current and prospective customers (d = -1.19, p = .007) in such a way that the latter group showed higher levels of SV than did the former group when they were exposed to an explicit as compared to an implicit message. Use of SV significantly decreased brand expertise (b = -.11, p = .006). The direct effect of language
forcefulness on brand expertise remained valid when controlled for SV and customer status (c’ = -.35, p = .004), despite the significant mediating effects of SV that were found for both customer types (PC = -.13, CI = [-.30, -.03]; CC = -.02, CI = [-.10, -.04]). The negative effect of explicit language as compared to implicit language on brand expertise via SV was stronger for prospective than for current customer.
Assertions of confidence. The CS-AoC-BE model is not significant (CI = [-.10, .02]). Explicit language did lead to significantly higher levels of AoC (M = 4.00, SD = 1.23) than did implicit language (a = 3.33, p = .008; M = 3.46, SD = 1.37), but engagement in this resistance strategy had no significant effect on brand expertise (b = .03, p = .550), nor was it influenced by customer status (d = -.48, p = .169). Current and prospective customers thus both engaged in AoC to a similar extent when exposed to an implicit or explicit message, but their use of this strategy did not alter their evaluations of brand expertise. AoC was also not found to be a significant mediator of the effect of language forcefulness on brand expertise for either customer group (PC = .02, CI = [-.04, .10]; CC = .01, CI = [-.01, .07]) and the direct effect of language forcefulness on brand expertise remained significant when controlled for AoC and customer status (c’ = -.42, p = .001).
Message derogation. The full CS-MD-BE model was not significant (CI = [-.03, .46]). Explicit language led to significantly higher levels of MD (a = 1.60, p < .000; M = 3.53, SD = 1.71) than implicit language (M = 2.41, SD = 1.23). Higher levels of SD in turn significantly decreased brand expertise (b = -.29, p < .000). Participants exposed to the explicit message thus shower higher levels of MD than participants exposed to the implicit message, an effect that did not vary between the two customer groups (d = -.65, p = .108). Higher levels of MD caused brand expertise
