The SAGE Handbook of Persuasion
Page 26
The relation between intention and behavior is also conditional on actual control over behavioral performance (Ajzen, 1985; Fishbein & Ajzen, 2010). People are thought to have actual control over behavioral performance when they have the necessary skills and when the situation does not impose constraints on behavioral performance. Thus, when despite positive intentions people do not perform a behavior, behavioral nonperformance is not a motivational problem but a problem of competence (i.e., deficient skills or abilities) and means (i.e., presence of environmental constraints). It is here where the aforementioned implementation intentions prove useful; actual behavior is more likely when people plan how and when to act on their intention (Norman & Conner, 2005; van Osch et al., 2009), possibly because planning requires people to consider the skills it takes and the obstacles they are up against when they would perform a particular behavior.
Attitude and Behavioral Beliefs
Attitude is an evaluation of performing a future behavior in terms of “favor or disfavor, good or bad, like or dislike” (Fishbein & Ajzen, 2010, p. 78). Although attitude is typically analyzed with a single composite scale, attitude is thought to have two aspects, namely an instrumental (or cognitive) aspect, indicated by perceptions of, for example, how foolish or wise, useful or useless performing a behavior is, and an experiential (or affective) aspect, indicated by how unpleasant or pleasant, unenjoyable or enjoyable performing the behavior is perceived to be. The relative importance of instrumental and experiential aspects of attitude as determinants of intention have clear implications for persuasive messages; if instrumental attitude matters most, a message should emphasize the usefulness of the recommended behavior, but if experiential attitude is more important, a message should emphasize how enjoyable the behavior is. Unfortunately, however, because published reports often do not make clear whether attitude was measured with instrumental, experiential, or both types of items, inferences about when instrumental and experiential attitude contribute to behavioral prediction cannot be made with full confidence. The question whether differential impact is predictable thus deserves more systematic inquiry than it has received thus far.
According to reasoned action theory, attitude formation is the process by which a potentially large set of specific beliefs, which has associated with a behavior over time, informs an overall sense of favorableness toward the behavior. Consistent with expectancy-value perspectives, attitude is a multiplicative combination of behavioral beliefs, which are perceptions of the likelihood that performing a particular behavior will have certain consequences, and an evaluation of those consequences in terms of good or bad. For example, two persons may both believe that if they use a tanning bed, they will get a tan. In addition, person A thinks that being tanned is good, but person B does not. In this single belief example, both person A and person B think that using a tanning bed will give them a tan, but because their opposite evaluations of being tanned person A’s attitude toward using a tanning bed is positive and person B’s attitude is negative. This makes clear that both beliefs about behavioral consequences and evaluations of those consequences need to be considered to determine favorableness toward a behavior. It also makes clear that to change attitude, persuasive messages can address beliefs about the likelihood of particular consequences of a behavior but also address evaluations of those consequences. For example, suppose that people already believe that unprotected sex may lead to gonorrhea but do not evaluate gonorrhea as a very serious disease. In this case, a message does not need to argue that unprotected sex can lead to gonorrhea, but can improve attitude toward using condoms if a message convinces that gonorrhea is quite serious.
Although belief-evaluation product terms have been found to correlate strongly with attitude (Albarracín et al., 2001), they typically do not explain much more variance in attitude than the separate behavioral beliefs (e.g., Armitage, Conner, Loach, & Willetts, 1999). For this reason, most investigators only assess behavioral beliefs, or the perceived likelihood of behavioral consequences. Note, however, that for statistical reasons product terms are unlikely to be associated with large effects in regression analysis, which is the method commonly used to test reasoned action (Ajzen & Fishbein, 2008; Yzer, 2007). We should be careful not to abandon conceptual ideas on the basis of empirical results if those results reflect statistical artifacts.
Perceived Norm and Normative Beliefs
To capture the influence of people’s social environment on their intention to perform a particular behavior, Fishbein and Ajzen (1975; Ajzen & Fishbein, 1973; Fishbein, 1967) proposed the concept of subjective norm as a second determinant of behavioral intention. In the theory of reasoned action (Fishbein & Ajzen, 1975) subjective norm is the extent to which I believe that other people think that I should or should not engage in a particular behavior. Other scholars refer to subjective norm as injunctive norm (Cialdini, Reno, & Kallgren, 1990), and in recent years, reasoned action theorists have used “injunctive norm” rather than “subjective norm” to indicate expected approval or disapproval from others (Fishbein, 2000; Fishbein & Ajzen, 2010).
The question whether subjective norm is able to capture all relevant perceived social influence has been controversial. This question in large part stemmed from empirical findings in which subjective norm contributed little to the explanation of intention (Albarracín et al., 2001; Cooke & French, 2008; Hagger et al., 2002). Note, however, that there is evidence that subjective norm matters in collectivistic populations (Giles, Liddell, & Bydawell, 2005; Lee & Green, 1991), in younger samples (Albarracín, Kumkale, & Johnson, 2004; van den Putte, 1993), and for behaviors that have salient social aspects (Cooke & French, 2008; Finlay, Trafimow, & Moroi, 1999), which implies that normative messages can have strong persuasive potential for some identified segments and behaviors. Even so, because much work found relatively small subjective norm effects, many investigators have tested alternative normative measures, including, among others, personal norm, verbal approval, social support, and descriptive norm (e.g., Larimer, Turner, Mallett, & Geisner, 2004; van den Putte, Yzer, & Brunsting, 2005).
In recognition of a need to expand the scope of the normative component, reasoned action theory currently posits a perceived norm component that is the composite of injunctive and descriptive norms (see also Fishbein, 2000). The descriptive norm indicates the extent to which I believe that other people perform a particular behavior themselves (Cialdini, Reno, & Kallgren, 1990). A meta-analysis of 14 correlations showed that descriptive norms explained variance in behavioral intention that subjective norms did not, supporting the discriminant validity of the descriptive norm variable (Rivis & Sheeran, 2003). In addition, injunctive and descriptive norms can have differential effects (Larimer et al., 2004), not only in magnitude but also in direction (Jacobson, Mortensen, & Cialdini, 2011). Thus, although in the context of reasoned action theory, injunctive and descriptive norms can be analyzed with a composite perceived norm scale, it may prove useful to also examine the effects of these variables separately.
Injunctive and descriptive norm measures tap normative perceptions regarding “most people who are important to me.” Perceived norm thus reflects perceived social pressure to perform or not to perform a behavior that is generalized across specific referents. It is a function of beliefs about particular individuals; whether particular individuals think I should perform a behavior (injunctive normative beliefs) or whether those individuals perform the behavior themselves (descriptive normative beliefs). However, believing that a particular individual prescribes a certain behavior will not matter if one does not care what that individual thinks, that is, if one is not motivated to comply with that individual. For example, someone affected by diabetes may expect that her doctor will approve her injecting insulin, but also believe that her friends will disapprove, or believe that her insulin-dependent friends do not self-inject. If it is more important for her to do what her peers want her to do than what her doctor wants her to do, then she will experience an overall sense o
f pressure against injecting insulin.
In more general terms, perceived norm is a function of normative beliefs about particular individuals weighed by the extent to which someone wants to comply with those individuals. However, as discussed in the context of multiplicative composites of behavioral beliefs and their evaluations, effects of product terms are hard to demonstrate in regression analysis. Reasoned action research often relies on regression analysis, which explains why there is not much evidence to support multiplicative composites of normative beliefs and motivation to comply (Fishbein & Ajzen, 2010). The usefulness of normative beliefs and motivation to comply should not be rejected if a lack of empirical support for these measures is caused by a statistical artifact. For example, Giles and colleagues (2005) examined both normative beliefs and motivation to comply regarding condom use in a sample of Zulu adults. Their analysis allowed them to identify important sources of influence, which in turn could inform decisions about who to target in behavior change interventions.
Perceived Behavioral Control and Control Beliefs
Concerned that the theory of reasoned action’s focus on volitional behavior unnecessarily restricted the scope of the theory, Ajzen (1985) argued that the theory could also predict nonvolitional behavior if it would address perceptions of control over behavioral performance. His inclusion of a perceived behavioral control variable as an additional determinant of intention and behavior established the theory of planned behavior (Ajzen, 1985, 1991). Perceived behavioral control was initially defined as “… people’s perception of the ease or difficulty of performing the behaviour of interest” (Ajzen, 1991, p. 183), and “compatible with … perceived self-efficacy” (p. 184). Consistent with this definition, items widely used to measure perceived behavioral control ask how much control people believe they have over performing a behavior, how easy or difficult they believe performing the behavior will be, or how confident they are that they can perform the behavior.
The proposed equivalence of perceived control, perceived difficulty, and self-efficacy has been the subject of considerable debate. Arguments in that debate for the most part are based on empirical tests of the dimensionality of perceived behavior control. A common finding from such tests is that confidence-framed items and control-framed items load onto separate factors (e.g., Armitage & Conner, 1999; Kraft, Rise, Sutton, & Røysamb, 2005). Importantly, these two factors are often interpreted as indicating “perceived behavioral control” and “self-efficacy,” suggesting a theoretical distinction between the two (Norman & Hoyle, 2004; Terry & O’Leary, 1995). Building on this idea, investigators have used the two item clusters to explore whether perceived behavioral control or self-efficacy offers a better explanation of intention or behavior (e.g., Pertl et al., 2010; Rodgers, Conner, & Murray, 2008).
The contention that perceived behavioral control and self-efficacy are theoretically distinct is unconvincing, however, if based solely on empirical criteria (such as proportions of variance explained) and without careful consideration of what these concepts are supposed to mean. For example, Terry and O’Leary (1995) purported to contrast perceived control and self-efficacy, but only used easy-difficult items to measure self-efficacy. It is not clear, however, why easy-difficult items are best seen as self-efficacy. Indeed, there is evidence that at least in some behavioral domains, easy-difficult is more closely related with attitude (Kraft et al., 2005; Yzer, Hennessy, & Fishbein, 2004) or intention (Rhodes & Courneya, 2003) than with control. Thus, whereas control items often load on two separate factors, this by itself does not irrefutably confirm the conceptual separation of perceived control and self-efficacy. Rhodes and Courneya (2003) warn in this regard against backward theorizing: “… items should be created to indicate theoretical concepts; theoretical concepts should not be created to indicate items!” (p. 80).
Fishbein and Ajzen (2010) similarly observe that “… although there is good empirical evidence that items meant to assess perceived behavioral control can be separated into two factors, identifying them as self-efficacy expectations and perceived control is misleading and unjustified” (p. 165). They argue that self-efficacy (Bandura, 1997) and perceived behavioral control are conceptually similar; both center on people’s perception of whether they can carry out a particular behavior. Consistent with this, reasoned action theory posits that perceived behavioral control/self-efficacy is a latent variable that has two aspects, namely capacity and autonomy. Capacity is indicated by items asking people how certain they are that they can perform a behavior. Autonomy is indicated by items asking people how much they feel that performing a behavior is up to them. Capacity and autonomy can be congruent, but there are situations in which they are not. For example, someone may believe that the decision to climb a tall building is up to him, but feel certain that he cannot do so because he is afraid of heights. Depending on the purpose of the investigation, capacity and autonomy thus can be combined or analyzed separately. Similarly, to enhance perceived behavioral control over a behavior, persuasive messages can focus on skill building, emphasize autonomous decision-making, or do both. The appeal of a multiaspect interpretation of perceived behavioral control is that it clarifies its conceptual definition, and refocuses our attention to the possibility of additive contributions of capacity and autonomy to behavioral prediction rather than superiority of one over the other. It also is a new idea, and thus should be a priority in future research.
The belief basis of perceived behavioral control consists of control beliefs (i.e., the perceived likelihood of having particular resources and opportunities for behavioral performance) and perceived power (i.e., the extent to which those resources and opportunities facilitate or obstruct behavioral performance). Perceived behavioral control is proposed to be the sum of the control beliefs-perceived power product terms. The belief basis of perceived behavior control has received curiously little research attention (see, e.g., Armitage & Conner, 2001). Therefore, and also considering the recent reconceptualization of perceived behavior control, systematic tests of control beliefs offer good opportunities for theoretical advancement.
Explaining Intention
Reviews of studies on determinants of intention have found multiple correlations in the R = .55−.70 range (e.g., Albarracín et al., 2001; Armitage & Conner, 2001; Hagger et al., 2002; Rivis & Sheeran, 2003; van den Putte, 1993). These results are impressive, particularly considering that they are based on studies that differ considerably in inclusion and measurement of predictor variables. At the same time, it should be noted that these multiple correlations reflect the effects of direct measures of attitude, perceived norm, and/or perceived behavioral control on intention. Relatively few studies have examined the role of beliefs in intention formation. Van den Putte (1993), for example, reports that of the 150 independent samples he analyzed, only 18 measured both behavioral beliefs and attitude, and only 13 measured both normative beliefs and subjective norm. The curious neglect of beliefs is disconcerting, because beliefs are the basis of persuasive messages that seek to change behavior.
A possible explanation for this phenomenon is that because of the availability of attitude, perceived norm, perceived behavioral control, and intention measure templates (e.g., Fishbein & Ajzen, 2010), designing measures of these four variables is a fairly straightforward affair. However, determining which beliefs are salient in a particular population is not as straightforward: “… although an investigator can sit in her or his office and develop measures of attitudes, perceived norms and [perceived behavioral control], she or he cannot tell you what a given population (or a given person) believes about performing a given behavior. Thus one must go to members of that population to identify salient outcome, normative and [control] beliefs” (Fishbein, 2000, p. 276). Recommendations for belief elicitation procedures are also available, however, (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 2010), and there thus is no good reason for disregarding beliefs if one seeks to explain intention.
Background Factors and the Question o
f Sufficiency
Beliefs originate from a large number of sources. Interaction with other people, engagement with media messages, growing up in a particular culture, membership of a religious community, and even gender and personality, for example, can all play a role in forming and shaping beliefs about a particular behavior. In the language of reasoned action theory, these variables are background factors, which are possibly but not necessarily related with beliefs. Similarly, background factors do not affect intention and behavior directly, but indirectly through beliefs. Thus, for example, if gender is empirically associated with intention or behavior, gender also should be correlated with beliefs, that is, men and women should hold different beliefs (Fishbein, 1967; Fishbein & Ajzen, 2010). Such findings can usefully inform decisions about which beliefs to target in different gender segments.
The conceptualization of background factors is directly relevant for a persistent debate on the question whether reasoned action variables are sufficient for explaining intention and behavior (for review, see Fishbein & Ajzen, 2010, chapter 9). Relevant for the present discussion of background factors is a substantial body of research that proposed an extension of the theory to better account for intention. Specifically, a number of different variables have been suggested as a fourth determinant variable in addition to attitude, perceived norm, and perceived behavioral control, including, among many others, gender, self-identity, and culture. Such research efforts are commendable to the extent that they promote theoretical development. However, many recommendations for extending reasoned action theory do not start from compelling conceptual arguments, but instead rely on empirical markers such as change in proportion of explained variance. The logic that if a particular variable explains variance in intention, it must be an important predictor has important statistical problems (Trafimow, 2004). A correlation between a particular variable and intention therefore does not conclusively prove that the variable is a predictor of intention and not a background factor.