Attitude Measurement Research Paper

This sample Attitude Measurement Research Paper is published for educational and informational purposes only. Free research papers are not written by our writers, they are contributed by users, so we are not responsible for the content of this free sample paper. If you want to buy a high quality research paper on any topic at affordable price please use custom research paper writing services.

Abstract

Because attitudes have been assumed to be predispositions for behaviors, consumer researchers have assumed that attitudes should predict behaviors. However, the success of predicting behaviors from attitudes largely depends on how reliably and validly attitudes are measured. The validity of attitude measures, in turn, depends on a variety of issues, the most important of which is correspondence of measurement. Other concerns are whether one is measuring attitudes toward objects or toward behaviors, whether one wishes to obtain implicit or explicit attitude measures, whether the attitude pertains to a single behavior or to repeated behaviors, and whether attitudes can be distinguished from other variables.

Outline

Introduction
Two Meanings of Reliability
Two Meanings of Validity
The Historical Problem: Attitudes Did Not Predict Behaviors
Factor Analysis
Attitudes Toward Objects or Behaviors and Behavioral Categories
Attitude Specificity, Behavioral Intentions, and Behaviors
Open Attitude Measures
Implicit Attitude Measures
Direct and Indirect Attitude Measures
Distinctions That Are Important for Attitude Measurement
Consumer Applications
Conclusion

1. Introduction

The vast majority of applied researchers who measure attitudes do so because they wish to predict or affect people’s behaviors. Health psychologists wish to increase healthy behaviors such as exercise, healthy eating, and use of condoms; consumer researchers wish to induce people to buy products; and politically motivated researchers wish to predict and affect people’s votes and financial contributions. Researchers usually assume that attitudes are an important determinant of behaviors, with the accompanying assumption that knowledge of people’s attitudes implies an ability to predict and control behaviors. To obtain such knowledge of people’s attitudes, however, it is absolutely crucial to be able to measure them reliably and validly—the main topic of this research paper.

2. Two Meanings Of Reliability

An attitude measure can be reliable in one of two general ways. To understand the simplest type of reliability, suppose that a set of participants completes an attitude measure at a particular time and then completes it again at a later time. The correlation between the two test-taking occasions is the attitude measure’s test–retest reliability. However, there is a second meaning of reliability in that the term can be used to denote internal consistency. To understand internal consistency, suppose that an attitude measure is composed of several scales. If these scales truly measure the participants’ attitudes, all of the scales should correlate with each other. Another way of saying that the scales correlate with each other is to say that they have a high degree of internal consistency. Although there are different formulas for calculating different kinds of internal consistency, all of these formulas require an assumption that all of the scales that compose the measure are equally good. Based on this assumption, all of the formulas imply that increasing the number of scales increases the internal consistency of the whole attitude measure. Because both test–retest reliability and internal consistency are usually considered to be prerequisites for validity, it is common practice to use several scales to measure attitudes.

3. Two Meanings Of Validity

How does one know whether a particular attitude measure is valid? To answer this question, one must first ask ‘‘valid for what?’’ An attitude measure can be valid if it predicts another variable that is considered desirable to predict such as behaviors and behavioral intentions (i.e., motivations to perform behaviors). For example, a measure of condom use attitudes that actually predicts whether people will intend to use condoms, or whether they actually use condoms, can be said to have predictive validity.

Another answer to the ‘‘valid for what?’’ question is that the measure must actually measure attitude and not something else. It is possible that a particular attitude measure might predict behavior, not because it measures attitude but rather because it measures something else that is correlated with behavior. Researchers who are concerned with this are forced to deal with the issue of how to define an attitude. Not surprisingly, various researchers have used different definitions, depending on their theories. From this perspective, the valid measurement of attitudes depends on the theories in which attitudes play an important role. Empirically obtained evidence in support of a particular theory also supports the validity of the attitude measurement as it is specified by that theory, whereas empirical evidence against a particular theory tends to disconfirm the validity of the attitude measurement as it is specified by that theory. When the theory in which the attitude construct is embedded is empirically supported, the attitude measure that is part of that theory is said to have construct validity.

4. The Historical Problem: Attitudes Did Not Predict Behaviors

Since the 1920s, researchers have assumed that attitudes are predispositions for behaviors. If this is true, it follows that attitudes should predict behaviors. However, the vast majority of researchers from the 1920s through the 1960s obtained correlations that were either low or not statistically significant. The consistent inability of researchers to predict behaviors from attitudes suggested one of two possibilities. First, perhaps attitudes really do not matter very much for predicting behaviors. Second, perhaps attitudes do matter, despite the lack of empirical support, but the attitude measures used in previous studies were not valid. Although many researchers argued for the first possibility, other researchers eventually achieved a good deal of success by assuming the second possibility. These latter researchers redefined the idea of attitude in two ways, both of which resulted in attitude measures with greater construct and predictive validity.

4.1. Attitude Accessibility

One view derives from the social cognition tradition, but it does not receive much attention here because it has not been used much by applied researchers. Fazio used the idea of attitude accessibility to explain low correlations between attitudes and behaviors. He assumed that attitudes affect behaviors, but only when they are easy to retrieve from memory. When attitudes are not accessible, there is no reason to believe that they would affect behavior. Therefore, researchers who subscribed to this view explained previous findings of low attitude– behavior correlations by assuming that they were due to a failure on the part of researchers to measure accessible attitudes. In support of the accessibility view, much research during the 1980s and 1990s indicated that accessible attitudes are more predictive of behaviors than are less accessible attitudes.

Nevertheless, applied researchers have, for the most part, ignored the accessibility view. The reason for this is not clear. One possible reason is that many applied researchers are not particularly knowledgeable about this view. Another possible reason is that researchers who have favored the accessibility view have not been clear about how this view can be applied to predict and control behaviors.

4.2. Principle of Correspondence

Fishbein proposed the principle of correspondence, which provided the other solution to the problem of low attitude–behavior correlations. His main assumption is that behaviors have four components: action, target, time, and context. It is easiest to understand this idea with an example. Suppose that a researcher wishes to predict whether people will give blood at the campus blood drive on Tuesday. The action is ‘‘give,’’ the target is ‘‘blood,’’ the time is ‘‘on Tuesday,’’ and the context is ‘‘at the campus blood drive.’’ If the attitude measure does not correspond with the behavior measure in regard to action, target, time, and context, it is invalid and a respectable attitude–behavior correlation is not likely to be obtained. For example, if a researcher measures people’s attitudes toward ‘‘giving blood,’’ there is no reason to expect that this attitude measure will successfully predict whether people will ‘‘give blood at the campus blood drive on Tuesday.’’ Rather, to predict this behavior, it is necessary to measure people’s attitudes toward ‘‘giving blood at the campus blood drive on Tuesday.’’ This is often done by having people make a check mark on a scale with various options such as the following:

I extremely like=quite like=slightly like=neutral=slightly dislike=quite dislike=extremely dislike giving blood at the campus blood drive on Tuesday.

(Whether people actually performed the behavior could be assessed in a variety of ways such as checking the lists of people who donated at the campus blood drive on Tuesday and simply asking people whether they had given blood at the campus blood drive on Tuesday.)

The correspondence view explained previously obtained low attitude–behavior correlations in a straightforward way. The reason for the low correlations is that the attitude measures used in those studies were not valid because they did not obey the principle of correspondence. Results from two types of research paradigms supported this view. First, researchers measured attitudes and behaviors (or behavioral intentions, which are often assumed to be direct precursors to behaviors) according to the principle of correspondence and obtained much higher correlations than those that researchers had been able to obtain previously. Second, some researchers manipulated the degree of correspondence of attitude and behavior (or behavioral intention) measures experimentally and obtained much higher correlations when there was a high degree of correspondence than when there was not. In combination, the findings provided a very convincing case for the importance of measuring attitudes in accordance with the principle of correspondence. Therefore, it is now common practice for applied researchers to measure attitudes in this way, and the result has been a dramatically improved ability to predict behavior in a variety of domains such as dieting, exercising, drinking, smoking, voting, using seat belts, using condoms, and getting screened for cervical cancer.

5. Factor Analysis

As was discussed earlier, increasing the number of scales can increase the internal consistency of the attitude measure. Consequently, many researchers do not stop at having one attitude scale and instead have several scales. For example, in addition to the like– dislike example given previously, participants could respond to scales that include pairs such as wise– foolish, beneficial–harmful, enjoyable–not enjoyable, good–bad, and pleasant–unpleasant. When multiple scales are used to measure an attitude, the mean of a participant’s responses to the scales is often taken as the representation of his or her attitude.

Depending on the behavior of interest, it may happen that not all of the scales are equally good for measuring attitude. It may even turn out that, for some behaviors, some of the scales will measure something other than attitude. To test this possibility, many researchers habitually submit the various scales to a factor analysis. Factor analysis is a statistical technique that reduces a large number of items down to a smaller number of underlying dimensions. For example, on intelligence quotient (IQ) tests, a large number of items may be reduced down to dimensions such as verbal ability and mathematical ability. As another example, on personality tests, hundreds of items have been factor analyzed and reduced to five basic factors of personality. Ideally, when various attitude scales are submitted to a factor analysis, one factor that represents attitude should result. Although one factor is often obtained, it sometimes happens that more than one factor is obtained. In this case, the researcher must determine which factor is the ‘‘true’’ attitude factor—a determination that depends on a variety of issues that are too complicated and numerous to discuss here. In addition, obtaining more than one factor can be interpreted to mean that there is more than one component to the attitude. In this case, the answer to the question of which factor represents attitude is that all of them may do so.

6. Attitudes Toward Objects Or Behaviors And Behavioral Categories

The principle of correspondence implies a distinction between attitudes toward objects and attitudes toward behaviors. Suppose that a researcher is interested in using advertising to reduce discrimination and wishes to measure relevant attitudes. Unfortunately, it is not clear what the relevant attitudes are. In general, researchers in this area have measured attitudes toward the groups of interest such as women, African Americans, Jews, handicapped people, and people with AIDS. However, the principle of correspondence suggests that if one wishes to predict discriminatory behavior, this is not going to work. This is because although a measure of attitudes toward a particular group includes the target component of the behavior to be predicted, it does not include the action, time, and context components. If one wishes to predict a particular discriminatory behavior, it is necessary to use an attitude measure that includes all of the components of the criterion behavior. At the very least, this implies that the attitude measure will have to include the action component, which makes it a measure of an attitude toward a behavior rather than a measure of an attitude toward an object. For example, an attitude toward ‘‘hiring women’’ is an attitude toward a behavior, whereas an attitude toward ‘‘women’’ is an attitude toward an object. For a more mundane example, one could measure an attitude toward ‘‘Skippy peanut butter’’ or toward ‘‘buying Skippy peanut butter’’; the former is an attitude toward an object, whereas the latter is an attitude toward a behavior.

An argument has been made against using the principle of correspondence in some situations. In brief, the argument is that researchers are sometimes interested in a large number of behaviors. For example, a researcher might be interested in a wide range of behaviors that discriminate against women and not just hiring behaviors. Obviously, an attitude measure toward a single behavior is likely to be inadequate for predicting a variety of discriminatory behaviors. One solution to this problem is for the researcher to measure attitudes toward all of the discriminatory behaviors that are of interest and to predict each discriminatory behavior from its corresponding attitude measure. Another solution is for the researcher to list a set of discriminatory behaviors of interest and then lump all of them under the general category of ‘‘discriminatory behaviors.’’ Participants can then give their attitudes toward the whole behavioral category, and these can then be used to predict the behaviors in that category. Although there is some evidence that either of these solutions may have some validity, there is insufficient evidence for a strong conclusion. However, there is strong evidence that attitudes toward behaviors are better predictors of behaviors than are attitudes toward objects.

7. Attitude Specificity, Behavioral Intentions, And Behaviors

Attitudes have been shown to be good predictors of behaviors, but they have also been shown to be good predictors of people’s intentions to perform behaviors. In addition, such behavioral intentions have often been shown to be good predictors of behaviors, and most researchers believe that intentions are proximate causes of behaviors. This combination of empirical findings and theorizing in the area, along with the practical point that behavioral intentions tend to be much easier to measure than are real behaviors, has resulted in the widespread use of behavioral intentions as a substitute for behaviors. This substitution has resulted in two controversies that have not yet been settled. First, there is a controversy about whether behavioral intentions are close enough to real behaviors to justify using the former as a substitute for the latter. A large part of this controversy is based on the issue of how well behavioral intentions predict behaviors. To the extent that intentions do a good job of predicting behaviors, the substitution would be supported, whereas a lack of prediction would support the reverse conclusion. In fact, results vary widely. In general (but there are exceptions), behavioral intentions do a better job of predicting behaviors when the intention and behavior measures conform to the principle of correspondence than when they do not. Thus, the substitution of behavioral intentions for behaviors is most acceptable when one uses correspondent measures.

The use of behavioral intentions as a criterion measure brings up a second controversy. To understand the underlying reason for the controversy, consider that all behaviors are performed with a target, an action, a time, and a context, meaning that the measured behavior will automatically have these four elements. Therefore, the trick to obeying the principle of correspondence is to make sure that the attitude measure specifies all of these elements in a way that matches the behavior measure. But matters change when attitudes are used to predict behavioral intentions rather than actual behaviors. A behavioral intention does not necessarily have a target, an action, a time, and a context, as is illustrated by the following example. Imagine the behavior of ‘‘buying a television set.’’ For a person to actually perform this behavior, he or she must buy a particular brand of television set, at a particular time, at a particular store or from a particular Web site. But for this person to intend to perform the behavior, many of these elements need not be specified. In the case of this example, the intention specifies a rather vague target (television set rather than a specific brand of television set) and an action (buy) but not a time or a context. Therefore, to have an attitude measure that is correspondent with the intention measure, it too should specify the target (in a similarly vague way) and the action but not a time or a context. Consequently, when one attempts to predict behavioral intentions from attitudes, it is possible to fully specify target, action, time, and context or to not fully specify these four elements. In either case, the principle of correspondence is obeyed so long as the degree of specification of the four elements is the same for both the attitude and behavioral intention measures. Does the degree of specification affect the size of the obtained correlation between the attitude and behavioral intention measures? Currently, there is insufficient empirical evidence to answer this question. Moreover, even for the prediction of actual behaviors, it is possible for the researcher to be uninterested in a particular element (e.g., the store at which the television set was bought). In this case, the behavior measure can ignore the uninteresting element or not ignore it, and the attitude measure can ignore it or not ignore it, and so long as the two measures match, the principle of correspondence is obeyed. Will the prediction of the behavior from the attitude measure be greater if the uninteresting element is specified? Again, there is insufficient empirical evidence to know. In sum, whether one predicts behavioral intentions or behaviors, whether one predicts clear-cut behaviors or behavioral categories, and whether one is interested in more of the elements (e.g., target, action, time, context) or fewer of them, there is insufficient research to determine whether it is better to have measures that are as specific as possible or not.

8. Open Attitude Measures

Although there is widespread agreement that it is crucial to measure attitudes in accordance with the principle of correspondence, it has not always been clear how to do this. In particular, it has not always been clear how to measure attitudes toward repeated behaviors in accordance with this principle. To see the problem, suppose that a researcher is interested in predicting exercise behaviors from attitudes toward exercising. If the researcher were interested in only one performance of an exercise behavior, it would be easy to obey the principle of correspondence as follows:

Attitude measure: I like/dislike to engage in vigorous physical activity at least one time during the month of October. (Participant makes a check mark on a scale.)

Behavior measure: I engaged in physical activity at least one time during the month of October. (Participant confirms or disconfirms that he or she performed the behavior.)

Unfortunately, this method of measuring attitudes and behaviors does not work well for repeated behaviors. Suppose that a researcher is interested in regular exercise. There is no clear dividing line between ‘‘regular’’ and ‘‘not regular’’ exercise; therefore, it is not satisfactory to arbitrarily specify a number of exercise occasions that qualifies as regular. Courneya proposed the idea of open attitude and behavior measures to address this problem. His idea was to have the participant, rather than the researcher, specify the number of behaviors. Although the use of open measures has not been tested exhaustively (and the tests have focused more on predicting behaviors from intentions than from attitudes), preliminary evidence suggests that it results in an increase in the prediction of behaviors compared with other methods. The following is an example of open measures:

Attitude measure: I like to engage in vigorous physical activity times during the month of October. Behavior measure: I engaged in vigorous physical activity times during the month of October.

To see why open measures often improve the prediction of behaviors from attitudes, consider an example. Suppose that a person believes that exercising 16 days during the month of October is ‘‘regular exercise,’’ whereas exercising 15 days or less is not. In addition, suppose that the person has a positive attitude toward regular exercise but that he or she exercised only 15 days during October. With traditional measures, the person’s behavior would seem to be inconsistent with his or her attitude; exercising only 15 days is inconsistent with a positive attitude toward regular exercise (16 days). In contrast, using open measures, it can easily be seen that exercising for 15 days is quite consistent with the person’s positive attitude toward exercising for 16 days.

9. Implicit Attitude Measures

Attitude researchers have often found it useful to distinguish between implicit and explicit attitude measures. Explicit attitude measures are usually characterized as being conscious, deliberative, and controllable, whereas implicit attitude measures are usually characterized as being unconscious, unintentionally activated, and not controllable. Whether the distinction between explicit and implicit measures should be considered to be a dichotomy or a continuum is not a settled issue, although many researchers who favor implicit measures seem to favor a dichotomous interpretation.

There are several examples of implicit attitude measures. One kind of implicit measure makes use of response latency, where participants respond to a computer-presented stimulus by pressing a key, and faster reaction times are interpreted to indicate a stronger attitude toward the stimulus. Other kinds of implicit measures include memory tasks, physiological measures (e.g., galvanic skin response, heart rate), and indirect questionnaires (where participants are not asked directly about the attitude object or behavior).

Implicit measures, when compared with explicit ones, have both advantages and disadvantages. Some advantages are as follows. First, if the issue of interest is a socially sensitive one, it may be difficult to get honest responses through explicit measures, and so implicit measures provide a way in which to avoid this difficulty. A finding that supports this argument is that implicit and explicit attitude measures have been found to be more highly correlated for socially insensitive attitude objects than for socially sensitive ones. Second, if the researcher is interested in unconscious attitudes, a case can be made that implicit measures are more valid than explicit ones. Supporting evidence indicates that implicit attitude measures tend to be more predictive of implicit measures of other variables than do explicit attitude measures. Third, it is possible that implicit measures are more direct than explicit ones. This is because explicit attitude measures must be funneled through the conscious processing system, whereas implicit ones do not.

On the other hand, implicit measures also have disadvantages. First, if one is interested in predicting behavior, it is reasonably clear how to use the principle of correspondence to obtain valid explicit attitude measures, whereas it is less clear how the principle of correspondence can be used for implicit attitude measures. Second, the reliability of implicit attitude measures has not been thoroughly investigated, and the preliminary evidence suggests that they have less test– retest reliability than do explicit measures. Because reliability is a precursor to validity, this preliminary evidence also suggests that implicit measures may also be less valid than explicit ones. Third, some of the same studies that suggest that implicit attitude measures are better than explicit ones for predicting implicit measures of other variables also suggest that implicit attitude measures work less well than explicit ones for predicting explicit measures of other variables (e.g., behavioral intentions). Finally, the presumed advantages for implicit attitude measures depend on assumptions about the nature of consciousness, cognitive resources, controllability, and others that, although supported in the literature, have not been proven beyond a reasonable doubt.

Once a researcher has decided to use an implicit measure such as response latency, there is a further issue of how to characterize the central tendency of the person’s responses to repeated exposures to the attitude object. Although many researchers have used mean scores, there are problems with this. For one, response latency distributions tend to be highly skewed in the direction of longer response latencies. This is because a large number of factors, such as blinking, distraction, and daydreaming, can increase response latencies. Consequently, mean scores are likely to present a distorted view of participants’ actual central tendencies.

Several solutions to this problem have been proposed. The simplest solution is to use median, rather than mean, response latencies as the attitude measure. Another solution is to perform a reciprocal transformation of the data to reduce their skewness [1/x, 1/(x + 1) if any of the response latencies are less than 1 second, where x is the raw response latency]. A third solution is to perform a logarithmic transformation, which has the consequence of bringing the tail involving slower latencies closer to the center of the distribution. An additional recent solution involves a transformation of all of the latency data to z scores, and then mean or median z scores can be used as the attitude measure. In sum, there are several ways in which to deal with the problem of skewed latency distributions, and the choice of which method to use depends on considerations that are too numerous and complex to describe here.

10. Direct And Indirect Attitude Measures

Several expectancy–value attitude theories were proposed during the 1950s and 1960s. According to these theories, attitudes are a function of people’s assumptions about the probability of various consequences arising from the performance of a behavior and evaluations of how good or bad those consequences are. Although the theories differ in the precise ways in which people are postulated to combine subjective probabilities and evaluations of consequences, they nevertheless have a common implication. Because attitudes are caused by a combination of beliefs about consequences and evaluations, this combination can be used as an indirect attitude measure. Consistent with expectancy–value theories, a large number of findings indicate that indirect and direct attitude measures are highly correlated, at least when the measures are created in accordance with the principle of correspondence. Consequently, many researchers have used indirect attitude measures, rather than direct ones, to predict other variables of interest, notably behavioral intentions and behaviors.

There has been some controversy about whether direct or indirect attitude measures are better predictors of behavioral intentions or behaviors. Although indirect measures have the advantage of specifying the subjective probabilities and evaluations of consequences that determine attitudes, there are both theoretical and empirical reasons to prefer direct attitude measures if one wishes to predict behavioral intentions or behaviors. The theoretical reason stems directly from the assumption that indirect measures assess variables (subjective probabilities and evaluations) that determine attitudes that, in turn, are a determinant of behavioral intentions or behaviors. According to this reasoning, attitudes are a more proximal cause of behavioral intentions or behaviors than are subjective probabilities and evaluations. Under the assumption that more proximal predictor variables work better than do less proximal ones, it follows that because direct attitude measures are assessing a variable (attitude) that is more proximal to behavioral intentions and behaviors than are indirect measures (which assess subjective probabilities and evaluations), direct measures should be a better predictor of behavioral intentions and behaviors than are indirect measures.

Numerous studies have provided tests of this theoretical reasoning. In the vast majority of cases, the reasoning has been supported; direct measures are generally superior to indirect ones for predicting behavioral intentions and behaviors. An exception is when the attitude measure is not in accord with the principle of correspondence.

11. Distinctions That Are Important For Attitude Measurement

With the problem of low attitude–behavior correlations having been solved by the principle of correspondence, recent researchers have changed their focus to three other problems that are now addressed. First, there is an important issue of whether attitudes are really an amalgamation of a cognitive (thinking) and affective (feeling) component. If there are separate cognitive and affective components, an implication is that each of these components should be measured separately, thereby improving the prediction of behaviors in a variety of applied domains. Second, there has been a great deal of controversy over whether attitudes should be measured separately from subjective norms. It was not until recently that this controversy was finally resolved. Finally, there are some attitudes that are not easily measured according to the principle of correspondence, and it might be necessary to make an adjustment. These issues are addressed in the following subsections.

11.1. The Separate Measurement of Cognition and Affect

Most researchers believe that there are both cognitive (thinking) and affective (feeling) components to attitudes. Three types of evidence support this belief. First, many researchers have used factor analysis to reduce a large number of attitude scales down to a smaller number of factors. In most cases, two factors result, with cognitive items loading on one factor and affective items loading on the other.

A second type of evidence comes from hierarchical regression analyses. Put simply, it is possible to consider the unique contribution of cognition to predicting general attitudes (or intentions) or to consider the unique contribution of affect. The results of these studies tend to indicate that, for any particular behavior, either cognition or affect will make a statistically significant unique contribution—a finding that should not occur regularly if attitude does not have these two components. An additional nicety of hierarchical regression analyses is that the results are often consistent with researchers’ intuitions. For example, cognition has been shown to be the more important contributor for what seems to be the cognitively controlled behavior of ‘‘studying over winter break,’’ whereas affect has been shown to be the more important contributor for what seems to be the affectively controlled behavior of ‘‘smoking cigarettes.’’

Finally, Trafimow and Sheeran have made use of the associative hypothesis. They assumed that people have different types of beliefs about the consequences of behaviors and that some beliefs are more cognitive, whereas other beliefs are more affective. Given this, people are assumed to compare cognitive beliefs with other cognitive beliefs in the interest of forming the cognitive component of an attitude, and people are assumed to compare affective beliefs with other affective beliefs in the interest of forming the affective component of an attitude. But when people compare cognitive beliefs with other cognitive beliefs, or compare affective beliefs with other affective beliefs, they form associations; people form associations between cognitive beliefs and other cognitive beliefs, or between affective beliefs and other affective beliefs, but not between cognitive beliefs and affective beliefs. Consequently, when people are later asked to write down their beliefs about a behavior, writing a cognitive belief should cue the retrieval of another cognitive belief, whereas writing an affective belief should cue the retrieval of another affective one, thereby causing people’s belief lists to be clustered by belief type. In contrast, if people do not distinguish between cognitive and affective beliefs (or between cognitive and affective components of attitudes), such clustering should not occur. In fact, such clustering is obtained, further supporting the distinction between cognitive and affective components of attitudes.

Empirical support for the distinction between cognitive and affective attitude components has led researchers to measure them separately to maximize the prediction of other variables such as behavioral intentions. Recent findings indicate that the prediction of behavioral intentions is significantly enhanced when the cognitive and affective attitude components are measured separately, and this finding has been replicated across a wide range of behavioral domains.

11.2. The Separate Measurement of Attitudes and Subjective Norms

Since the 1960s, researchers have assumed that attitudes and subjective norms are different causes of behaviors (or, more often, behavioral intentions). Whereas attitudes have been assumed to be caused by beliefs about the personal consequences of performing a behavior, subjective norms have been assumed to be caused by beliefs about what important others think one should do. Because attitudes and subjective norms have been assumed to be different causes of behavioral intentions, it made sense to measure them separately to maximize the prediction of behavioral intentions. However, during the 1970s and 1980s, many researchers questioned the distinction between attitudes and subjective norms and argued that these were really different names for the same underlying idea. Obviously, if attitudes and subjective norms are different names for the same underlying construct, there is no reason to have distinct measures of each of them. Thus, there is an important measurement issue at stake in this controversy: Should attitudes be measured separately from subjective norms or not?

The argument against the distinction between attitudes and subjective norms was based on three issues. First, attitudes and subjective norms were often found to be highly correlated with each other, consistent with the notion that they are merely different names for the same underlying construct. Second, path analyses have sometimes indicated ‘‘crossover’’ effects, whereby attitudes and subjective norms affect each other. Third, philosophical arguments have been made that beliefs about consequences (i.e., behavioral beliefs) that are presumed to cause attitudes are not different from beliefs about the opinions of important others (i.e., normative beliefs) that are presumed to cause subjective norms. An example should make this issue clear. Suppose that someone has the behavioral belief that ‘‘my father will disagree if I eat chocolate’’ or the normative belief that ‘‘my father thinks I should not eat chocolate.’’ Many researchers have argued that these two statements are just different ways of saying the same thing; therefore, if the cause of attitudes and subjective norms is the same, attitudes and subjective norms must also be the same.

Five kinds of findings have resolved most of the disagreement in favor of the distinction between attitudes and subjective norms. First, in general, behavioral beliefs have been found to be more highly correlated with attitudes than with subjective norms, and normative beliefs have been found to be more highly correlated with subjective norms than with attitudes—precisely what one would expect if behavioral beliefs cause attitudes and normative beliefs cause subjective norms. Second, although attitudes and subjective norms sometimes have been found to be highly correlated, the correlation often has been found to be low or moderate. Third, attitudes and subjective norms have been manipulated experimentally, with different effects on behavioral intentions pertaining to different types of behaviors; manipulating attitudes has a larger effect on some behaviors, whereas manipulating subjective norms has a larger effect on others. Fourth, behavioral beliefs have been shown to be more strongly associated with other behavioral beliefs than with normative ones, and normative beliefs have been shown to be more strongly associated with other normative beliefs than with behavioral ones. Clearly, regardless of philosophical validity of the distinction between the two types of beliefs, people do make the distinction, contradicting the argument that attitudes and subjective norms have the same cause. Finally, there are individual differences between people in that some people are more under attitudinal control across a large number of behaviors, whereas others are more under normative control. If attitudes and subjective norms were the same thing, reliable individual differences in attitudinal or normative control should not be obtained. In sum, there is a great deal of evidence that attitudes and subjective norms are different from each other, and they should be measured separately.

12. Consumer Applications

Attitude measurement is of more than just theoretical interest. Consumer and marketing researchers have applied the forgoing principles of attitude measurement in a variety of domains such as substance abuse among adolescents and tobacco use among college athletes as well as trying a new diet suppressant, getting mammograms, using condoms, drinking and driving, eliciting donations, purchasing environmentally friendly products, paying more for energy from renewable sources, purchasing food, voting, purchasing software, and even purchasing attitude research (by marketing directors). In these domains, as well as in many other domains, improved attitude measurement has resulted in an improvement in the prediction of behavioral intentions and behaviors from attitudes.

Improving the prediction of behavioral intentions and behaviors is not the only function of good attitude measures. It sometimes happens, although not very often, that an attitude does not do a good job of predicting a particular behavior. Before psychologists and consumer researchers had valid attitude measures, a low correlation between an attitude and a behavior was susceptible to at least two explanations. First, the low correlation could show that the behavior is caused by something other than an attitude. Second, the low correlation could be due to an invalid attitude measure. An advantage of valid attitude measures is that they decrease the plausibility of the latter explanation and thereby increase the plausibility of the former one. Thus, when valid attitude measures nevertheless result in low correlations between attitudes and behaviors, researchers can be more confident in exploring other variables. A well-researched example of such an area is condom use. Although attitudes are capable of predicting condom use to some degree, other variables have also been shown to be good (perhaps better) predictors. Two of these are subjective norms and confidence that one knows whether others (e.g., one’s sexual partner) think a condom should be used.

To illustrate the importance of valid attitude measures for consumer research regardless of whether the measures show that attitudes are a strong or weak predictor of behaviors, consider an example of a consumer researcher who wishes to increase sales of a particular product. Before investing money in an ad, the researcher needs to know what variables to address in the ad. If the behavior of buying the product is under attitudinal control for the population of interest, it makes sense for the ad to focus on variables that are likely to affect people’s attitudes toward buying the product. Some of these variables might be beliefs about the product and affect. But what if the behavior is not under attitudinal control? In that case, there is little point in focusing an ad on variables designed to affect attitudes; for a behavior that is not under attitudinal control, there is no reason to believe that causing attitude change will increase sales. Consequently, the consumer researcher would be better served by creating an ad that focused on a different variable such as subjective norms. In general, having valid attitude measures increases the confidence that researchers can have in their data and provides a more solid basis for creating ads. Data showing that attitudes are good predictors of the behavior of concern provide a strong reason for creating an ad that is designed to affect attitudes; otherwise, the data provide a strong reason for creating an ad that focuses on other variables.

A further issue in attitude measurement, as applied to evaluating the effects of advertising, concerns the types of attitudes that should be measured. Consumer researchers have measured the effects of an ad on attitudes toward the ad, the product, or the brand. It has been rarer for researchers to measure the effects of an ad on attitudes toward the behavior the ad was designed to influence (e.g., buying the advertised product, voting for the advertised candidate). According to the principle of correspondence, if the reason for creating an ad is to influence behavior, it is precisely this last type of attitude that should be measured. Although there have been some demonstrations that an ad affects attitudes toward the behavior rather than merely attitudes toward the ad, the product, or the brand, more research is needed to establish the size of these effects. This is particularly so because the effects are likely to depend on a large number of variables such as the product domain, the type of ad, the type of use to which the product is put, and the frequency with which people are exposed to the ad.

One final example illustrates the importance of whether one measures attitudes toward the ad, the brand, the product, or the behavior of buying the product. Suppose that a consumer researcher for a company wishes to increase sales of a particular product. In addition, suppose that this researcher is evaluating an ad that has been shown to cause a change in people’s attitudes toward the ad, the brand, or the product. It should be clear from the principle of correspondence that these attitudes, despite their seeming importance, are likely to not be particularly relevant to whether people will buy the product.

There may be cases where the consumer researcher’s goal is something other than increasing sales of a particular product. Perhaps the goal is to increase sales of all of the products made by the company. In that case, it might be worthwhile to run an ad that focuses on increasing people’s attitudes toward the brand. Even though such an ad might be unlikely to cause much of an increase in the sales of a particular product, a small increase in the sales of several products may justify the cost of the ad.

13. Conclusion

There has been a great deal of progress in how attitudes are measured. Current attitude measures are more reliable, more valid, and more correspondent to behaviors than were attitude measures in the past. In addition, more is known about some of the relevant issues that underlie attitude measurement. For example, recent research indicates that cognition and affect are both components of attitudes and should be measured separately. Recent research also indicates that attitudes are different from subjective norms and that the two variables should be measured separately. The consequence of such progress is that the prediction of behavioral intentions and behaviors from attitudes is much greater than it has ever been before, and this has been demonstrated in a wide variety of domains.

References:

Ajzen, I. (1988). Attitudes, personality and behavior. Chicago: Dorsey.
Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice Hall.
Courneya, K. S. (1994). Predicting repeated behavior from intention: The issue of scale correspondence. Journal of Applied Social Psychology, 24, 580–594.
Dovidio, J. F., Kawakami, K., & Beach, K. R. (2001). Implicit and explicit attitudes: Examination of the relationship between measures of intergroup bias. In R. Brown, & S. Gaertner (Eds.), Blackwell handbook of social psychology: Intergroup processes (pp. 175–197). Malden, MA: Blackwell.
Fazio, R. H. (1990a). Multiple processes by which attitudes guide behavior: The MODE model as an integrative framework. In M. P. Zanna (Ed.), Advances in experimental social psychology (Vol. 23, pp. 75–109). San Diego: Academic Press.
Fazio, R. H. (1990b). A practical guide to the use of response latency in social psychological research. In C. Hendrick, & M. S. Clark (Eds.), Review of personality and social psychology, Vol. 2. Research methods in personality and social psychology (pp. 74–97). Newbury Park, CA: Sage.
Fishbein, M. (1980). Theory of reasoned action: Some applications and implications. In H. Howe, & M. Page (Eds.), Nebraska Symposium on Motivation, 1979 (pp. 65–116). Lincoln: University of Nebraska Press.
Trafimow, D. (1998). Attitudinal and normative processes in health behavior. Psychology and Health, 13, 307–317.
Trafimow, D., & Sheeran, P. (1998). Some tests of the distinction between cognitive and affective beliefs. Journal of Experimental Social Psychology, 34, 378–397.