TRA asserts that the most important determinant of behavior is behavioral inten- tion (see unshaded boxes in Figure 4.1). Direct determinants of individuals’ behav- ioral intention are their attitude toward performing the behavior and their subjective norm associated with the behavior. TPB adds perceived control over the behavior, taking into account situations where one may not have complete volitional control over a behavior (see shaded boxes in Figure 4.1).
Normative beliefs
Motivation to comply
Intention to perform the behavior Subjective
norm Behavior
Control beliefs
Perceived power
Perceived control Behavioral
beliefs
Evaluations of behavioral
outcomes
Other individual difference variables Personality
traits Attitudes
towards targets Demographic
variables External
variables Attitude
FIGURE 4.1.
Theory of Reasoned Action and Theory of Planned Behavior.**Note:Upper light area shows the Theory of Reasoned Action; entire figure shows the Theory of Planned Behavior.
Attitude is determined by the individual’s beliefs about outcomes or attributes of performing the behavior (behavioral beliefs), weighted by evaluations of those out- comes or attributes. Thus, a person who holds strong beliefs that positively valued outcomes will result from performing the behavior will have a positive attitude to- ward the behavior. Conversely, a person who holds strong beliefs that negatively val- ued outcomes will result from the behavior will have a negative attitude.
Similarly, a person’s subjective norm is determined by his or her normative be- liefs, that is, whether important referent individuals approve or disapprove of per- forming the behavior, weighted by his or her motivation to comply with those referents.
A person who believes that certain referents think she should perform a behavior and is motivated to meet expectations of those referents will hold a positive subjective norm. Conversely, a person who believes these referents think she should not perform the behavior will have a negative subjective norm, and a person who is less motivated to comply with those referents will have a relatively neutral subjective norm.
TRA assumes that the most important direct determinant of behavior is behav- ioral intention. Success of the theory in explaining behavior depends on the degree to which the behavior is under volitional control (that is, individuals can exercise a large degree of control over the behavior). It is not clear that the TRA components are sufficient to predict behaviors in which volitional control is reduced. Thus, Ajzen and colleagues (Ajzen, 1991; Ajzen and Driver, 1991; Ajzen and Madden, 1986) added perceived behavioral control to TRA to account for factors outside individual con- trol that may affect intentions and behaviors. With this addition, they created the The- ory of Planned Behavior (TPB; see shaded boxes in Figure 4.1). Perceived control is determined by control beliefs concerning the presence or absence of facilitators and barriers to behavioral performance, weighted by their perceived power or the im- pact of each control factor to facilitate or inhibit the behavior.
Ajzen’s inclusion of perceived control (Ajzen, 1991) was based in part on the idea that behavioral performance is determined jointly by motivation (intention) and abil- ity (behavioral control). A person’s perception of control over behavioral perform- ance, together with intention, is expected to have a direct effect on behavior, particularly when perceived control is an accurate assessment of actual control over the behav- ior and when volitional control is not high. The effect of perceived control declines, and intention is a sufficient behavioral predictor in situations in which volitional con- trol over the behavior is high (Madden, Ellen, and Ajzen, 1992). Thus, similar to Triandis’s (1980) conceptualization of facilitating conditions, perceived control is ex- pected to moderate the effect of intention on behavior. However, this interaction hy- pothesis has received very little empirical support (Ajzen, 1991; Yzer, 2007).
TPB also postulates that perceived control is an independent determinant of be- havioral intention, along with attitude toward the behavior and subjective norm. Hold- ing attitude and subjective norm constant, a person’s perception of the ease or difficulty of behavioral performance will affect his behavioral intention. Relative weights of these three factors in determining intentions should vary for different behaviors and populations. Few studies have operationalized perceived control using the underly- ing measures of control beliefs and perceived power; instead, researchers have mostly used the direct measure of perceived control (Ajzen, 2002).
TRA and TPB assume a causal chain that links behavioral beliefs, normative be- liefs, and control beliefs to behavioral intentions and behaviors via attitudes, subjective norms, and perceived control. Hypothesized causal relationships among model com- ponents are clearly specified, and measurement and computation are delineated by Ajzen and Fishbein (Ajzen and Fishbein, 1980; Ajzen, 1991; Ajzen, 2006). This is one of the major strengths of the TRA/TPB approach. Other factors, including demo- graphic and environmental characteristics, are assumed to operate through model con- structs and do not independently contribute to explain the likelihood of performing a behavior.
Measures of TRA and TPB Constructs
TRA and TPB measures can use either 5- or 7-point scales. A person’s behavioral be- liefs about the likelihood that performance of the behavior will result in certain out- comes are measured on bipolar “unlikely-likely” or “disagree-agree” scales. Evaluations of each outcome are measured on bipolar “good-bad” scales. For example, one out- come of “my quitting smoking” may be that this “will cause me to gain weight.” A person’s behavioral belief about this outcome is measured by having him rate the like- lihood that “my quitting smoking will cause me to gain weight.” The person’s eval- uation of this outcome is measured by having him rate the degree to which “my gaining weight” is good versus bad. These behavioral belief and evaluation ratings are usu- ally scored from −3 to +3, capturing the psychology of double negatives, where a be- lief that a behavior will not result in a negative outcome contributes positively to the person’s attitude. An “indirect measure” of the person’s attitude toward per- forming the behavior is computed by first multiplying her behavioral belief concern- ing each outcome by her corresponding outcome evaluation ratings and then summing these product scores across all outcomes of the behavior.
In the example, a person may believe that “quitting smoking” is very unlikely to result in “gaining weight” (belief scored as −3), and may evaluate gaining weight as very bad (evaluation scored as −3), resulting in a belief-evaluation product score of +9. Thus, the strong belief that performing the behavior will not result in (will avoid) a negatively valued outcome contributes just as positively to the person’s attitude as would a strong belief that the action will result (+3) in a positively valued (+3) out- come (product = +9). Conversely, a strong belief that the behavior will not result (−3) in a positively valued outcome (+3) contributes negatively (product = −9) to the per- son’s attitude, because performance of the behavior will not achieve a highly valued outcome. In the example of “quitting smoking,” beliefs and evaluations of all salient outcomes of this behavior will enter into the computation of an indirect measure of the person’s attitude.
Similarly, a person’s normative beliefs about whether each referent thinks he should perform the behavior are measured on bipolar scales scored −3 to +3, while the person’s motivation to comply with each referent is measured on unipolar scales scored 1 to 7. For example, one potential referent with regard to “quitting smoking”
might be the person’s best friend. A person’s normative belief concerning his best friend is measured by asking him to rate the degree to which he believes his best friend
thinks he should versus should not quit smoking. Motivation to comply is measured by having the person rate his agreement versus disagreement with the statement: “Gen- erally, I want to do what my best friend thinks I should do.” An indirect measure of the person’s subjective norm is computed by multiplying his normative belief about each referent by his motivation to comply with that referent and then summing these product scores across all referents.
Applications of TPB suggest that control beliefs regarding each factor should be measured on a bipolar likelihood of occurrence scale scored −3 to +3. Perceived power of each factor is measured on a bipolar “easy-difficult” scale (Terry, Gallois, and McCamish, 1993; Ajzen, 1991). For example, some individuals might identify
“restaurant smoking restrictions” as a factor that affects their perceived behavioral control over quitting smoking. A person’s control belief regarding this factor is meas- ured by having her rate her likelihood of encountering “a restaurant smoking restric- tion,” while perceived power is measured by having the person rate her perception of the effect of “restaurant smoking restrictions” in making it easier versus more diffi- cult to quit smoking. These measures are obtained for all factors identified as facili- tating or impeding the behavior. An “indirect measure” of the person’s perceived behavioral control is then computed by multiplying each control belief by the corre- sponding perceived power (impact) rating, and then summing these product scores across all control factors (Ajzen and Driver, 1991).
In addition to the indirect measures computed from behavioral, normative, and control beliefs, it is important to obtain a “direct measure” of each model component.
Table 4.1 summarizes the direct and indirect measures of attitudes, subjective norms, and perceived behavioral control. A direct measure of attitude toward performing the behavior is obtained using semantic differential scale items, such as “good-bad”
and “pleasant-unpleasant,” and summing them. A direct measure of subjective norm uses a single item, asking the person to rate “Most people important to me think I should” perform the behavior. This rating is made on a bipolar “unlikely-likely” or
“agree-disagree” scale. The direct measure of perceived behavioral control generally uses semantic differential scale items such as “under my control–not under my con- trol” and “easy-difficult.”
These direct measures are important for two reasons. First, direct measures are usually more strongly associated with intentions and behaviors than indirect meas- ures. The associations between the “direct” measures and behavioral intention indi- cate the relative importance of attitude, subjective norm, and perceived control in explaining or predicting a given behavior. It is important to demonstrate these asso- ciations before analyzing indirect measures. Second, indirect measures should be as- sociated strongly with direct measures to be assured that appropriate beliefs were included in the indirect measures and that the composite beliefs (behavioral, norma- tive, and control) are adequate measures of respective TRA/TPB constructs. Once this is demonstrated, indirect measures are of most interest. Behavioral, normative, and control beliefs help us understand what drives behaviors and provide a focus for intervention messages (von Haeften, Fishbein, Kasprzyk, and Montaño, 2001; Fish- bein and Cappella, 2006). The process of assessing which construct is most closely related to behavioral intention and deciding which behavioral, normative, and control
TABLE 4.1.
TRA, TPB, and IBM Constructs and Definitions.Construct Definition Measure
Behavioral Intention
Experiential Attitude (Affect) Direct Measure:
Indirect Measure:
Behavioral belief
Instrumental Attitude Direct Measure:
Indirect Measure:
Behavioral belief
Evaluation
Subjective (Injunctive) Norm Direct Measure:
Indirect Measure:
Normative belief Motivation to comply
Descriptive Norm Direct Measure:
Indirect Measure:
Normative belief
Perceived likelihood of per- forming the behavior
Overall affective evaluation of the behavior
Belief that behavioral per- formance is associated with certain positive or negative feelings
Overall evaluation of the behavior
Belief that behavioral performance is associated with certain attributes or outcomes
Value attached to a behav- ioral outcome or attribute
Belief about whether most people approve or disap- prove of the behavior Belief about whether each referent approves or disap- proves of the behavior Motivation to do what each referent thinks
Belief about whether most people perform the behavior Belief about whether each referent performs the behavior
Bipolar unlikely-likely scale;
scored −3 to +3
Semantic differential scales: for example, pleasant-unpleasant;
enjoyable-unenjoyable Bipolar unlikely-likely scale;
scored −3 to +3
Semantic differential scales:
for example, good-bad; wise- foolish
Bipolar unlikely-likely scale;
scored −3 to +3
Bipolar bad-good scale; scored
−3 to +3
Bipolar disagree-agree scale;
scored −3 to +3
Bipolar disagree-agree scale;
scored −3 to +3
Unipolar unlikely-likely scale;
scored 1 to 7
Bipolar disagree-agree scale;
scored −3 to +3
Bipolar disagree-agree scale;
scored −3 to +3
AttitudePerceived Norm
beliefs should be used to focus intervention messages is illustrated in the theory ap- plication example that follows.
Research Designs and Analytical Approaches to Testing TRA/TPB
A prospective study design is recommended to discern relationships between con- structs, with attitudes, subjective norms, perceived control, and intentions measured at one time point and behavior measured following a time interval. Cross-sectional studies are often used to test the TRA/TPB, but they may provide poor prediction and understanding of previous behavior because the time order of motivations and behav- ior cannot be discerned. Regression and structural equation analytic methods are usu- ally used to test relationships in the TRA/TPB (Rhodes and others, 2007; Bryan, Schmiege, and Broaddus, 2007). Relative weights of model constructs are determined empirically for the particular behavior and population under investigation. This in- formation provides guidance as to which constructs are most important to target for behavior change effort. Some behaviors are entirely under attitudinal control (Albar- racin and others, 2003), while others are under normative control (Albarracin, Kumkale, and Johnson, 2004; Durantini and others, 2006) or perceived control (Albarracin and
TABLE 4.1.
TRA, TPB, and IBM Constructs and Definitions, Cont’d.Construct Definition Measure
Perceived Behavioral Control Direct Measure:
Indirect Measure:
Control belief
Perceived power
Self-Efficacy Direct Measure:
Indirect Measure:
Self-efficacy belief
Overall measure of per- ceived control over the behavior
Perceived likelihood of oc- currence of each facilitating or constraining condition Perceived effect of each condition in making behav- ioral performance difficult or easy
Overall measure of ability to perform behavior Perceived ability to over- come each facilitating or constraining condition
Semantic differential scales: for example, under my control–not under my control; easy-difficult Unlikely-likely scale; scored −3 to +3 or 1 to 7
Bipolar difficult-easy scale;
scored −3 to +3
Certain I could not–certain I could scale for overall behavior;
scored −3 to +3 or 1 to 7 Certain I could not–certain I could scale; scored −3 to +3 or 1 to 7
Note:TRA/TPB constructs are shaded.
Personal Agency
others, 2005; Yzer, 2007). For example, in a study of adults over age forty, McLallen and Fishbein found colonoscopy intention to be almost completely under normative control, whereas exercise intention was influenced by both attitudes and perceived control (Fishbein and Cappella, 2006). Similarly, a behavior may be under attitudi- nal control in one population but under normative control in another population (Fish- bein, 1990, Fishbein, von Haeften, and Appleyard, 2001). Our research found that condom use with a main partner is primarily under normative control for female in- jecting drug users but influenced by attitude, norm, and perceived control for females who do not inject drugs (von Haeften and Kenski, 2001; Kenski and others, 2001).
Once the significant constructs are identified, analyses of the beliefs underlying those constructs can determine which specific behavioral, normative, or control beliefs are most strongly associated with intention and behavior, thus providing empirically iden- tified targets for intervention efforts.
Uses for and Evidence to Support TRA/TPB
The name Theory of Reasoned Action has often led to the misrepresentation that the focus is purely on “rational behavior” (for example, St. Lawrence and Fortenberry, 2007). This is far from correct. A fundamental assumption of TRA is that individu- als are “rational actors” who process information and that underlying reasons deter- mine motivation to perform a behavior. These reasons, made up of a person’s behavioral, normative, and control beliefs, determine his attitudes, subjective norms, and per- ceived control, regardless of whether those beliefs are rational, logical, or correct by some objective standard. (See Fishbein, 2007, for additional discussion regarding this aspect of the TRA/TPB.) A strength of TRA/TPB is that they provide a frame- work to discern those reasons and to decipher individuals’ actions by identifying, measuring, and combining beliefs relevant to individuals or groups, allowing us to understand their own reasons that motivate the behavior of interest. TRA and TPB do not specify particular beliefs about behavioral outcomes, normative referents, or con- trol beliefs that should be measured. As noted in the examples, relevant behavioral outcomes, referents, and control beliefs will likely be different for different popula- tions and behaviors.
TRA and TPB provide a framework to identify key behavioral, normative, and control beliefs affecting behaviors. Interventions can then be designed to target and change these beliefs or the value placed on them, thereby affecting attitude, subjec- tive norm, or perceived control and leading to changes in intentions and behaviors.
TRA/TPB has been applied to explain a variety of health behaviors, including ex- ercise, smoking and drug use, HIV/STD-prevention behaviors, mammography use, clinicians’ recommendation of and provision of preventive services, and oral hygiene behaviors. These studies generally have supported perceived control as a direct pre- dictor of both intentions and behaviors (Albarracin, Johnson, Fishbein, and Mueller- leile, 2001; Ajzen, 1991; Blue, 1995; Craig, Goldberg, and Dietz, 1996; Godin and Kok, 1996; Millstein, 1996; Montaño, Phillips, and Kasprzyk, 2000; Montaño, Thomp- son, Taylor, and Mahloch, 1997). However, most studies have used direct measures of perceived control, rather than computing perceived control from measures of con-
trol beliefs and perceived power concerning specific facilitators and constraints. The few studies that have measured control beliefs (indirect measure) found them to be important predictors of intentions and behaviors (Ajzen and Driver, 1991; Kasprzyk, Montaño, and Fishbein, 1998). Clearly, if perceived behavioral control is an impor- tant determinant of intentions or behaviors, knowledge of the effects of control be- liefs concerning each facilitator or constraint would be useful in the development of interventions.