WK2

CHAPTER 6 ATTITUDES AND PERSUASION: CHANGING HEARTS AND MINDS

The world's largest completed industrial merger to that date, between Daimler-Benz, Germany's largest company, and Chrysler Corporation, was announced May 7, 1998. The companies spent $25 million on public relations targeted to the merger alone. On the day of the announcement, across four continents, they telephoned 147 world leaders; held news conferences for 502 journalists; briefed 950 industry analysts; beamed 5,000 videos to television; phoned 10,000 union, political, and civic leaders; sent 140,000 e-mails; delivered letters and gifts to each of their 428,000 employees; and popped countless champagne corks. Nearly a year later, the head of Daimler-Chrysler reported that he was spending 40% to 50% of his time networking, still trying to convince his many audiences that the merger was a good idea. What were the issues, and why was it so urgent to persuade all these constituencies?

“Controlling and shaping a corporate image creeps into nearly every corner of a chief executive's day” (Ibrahim, 1999, p. C7). CEO Jürgen Schrempp's speaking engagements were booked more than a year in advance, including such venues as the World Economic Forum, industry executive clubs, and universities. In this merger, three top PR firms identified the major issues: American Jews' reactions to Daimler's use of World War II slave labor; Chrysler's diversity commitment, allegedly less a concern to Germans; and employee fears that the company would become all German or all American. These issues could have broken the deal, sparked strikes, lowered morale, tainted the product, killed the managers' careers, or ruined the merged company's stock. Regardless of all this effort, ten years later, Daimler had to sell its plummeting Chrysler division. Within a year, Daimler turned a tidy profit, while Chrysler tapped a $2 billion credit line, scrambling to redefine its image, tuning in to changing American economic conditions. Globalization has intensified the need for finely honed persuasive skills suited to multiple audiences, across multiple media.

At the start of the 21st century, each of students' currently most popular career choices—business, health, law, communications, politics, education, art, and science—requires the science of persuasion. Business people have to pitch their product; health practitioners have to induce patient compliance; lawyers have to argue their case; communications media have to win audiences; politicians have to impress their constituencies; educators have to sway their students; artists have to intrigue their viewers; and scientists have to convince their colleagues. Understanding attitudes and persuasion propels almost any professional career, even absent an unprecedented global industrial merger. Social psychologists have much to offer on this score, having studied attitudes from the outset of the field a century ago.

Attitudes have both conceptual and operational definitions, of course, and attitudes' functions fit some of the familiar core social motives, which appeared in the earliest social science of attitudes. As this chapter shows, the life course of an attitude includes, first, how it forms, primarily as a rapid, affect-based, perhaps unconscious understanding of what to approach or avoid; second, how it may change via pressures to understand in cognitively consistent ways or to understand and respond to a persuasive communication; and how attitudes may or may not affect behavior.

WHAT ARE ATTITUDES? THE IMPORTANCE OF BEING PERSUASIVE

Conceptual Definitions

At a minimum, attitudes entail evaluation. Experts agree that the sine qua non (“without which, nothing”) of an attitude is the positive or negative judgment of an attitude object (i.e., the entity about which one bears an attitude). Attitudes link an object to an evaluation (Albarracín & Vargas, 2010; Crano & Prislin, 2006; Ferguson & Fukukura, 2012). “Evaluating a particular entity with some degree of favor or disfavor” captures the flavor for Alice Eagly and Shelly Chaiken (1998, p. 269; Eagly & Chaiken, 2007), while Richard Petty and Duane Wegener (1998, p. 323) describe an “overall evaluation of persons (including oneself), objects, and issues.”

The focus on evaluation notwithstanding, attitudes can elicit three types of responses that reflect the more general psychological tripod within social psychology ( Chapter 1 ): affect, behavior, and cognition. The affective correlates of response comprise the emotions and feelings the attitude object elicits (whether “Never!”, “Yuck!”, “Hmmm … ”, “Wow!”, or “Yes! Yes!”). The cognitive correlates correspond to the beliefs the object elicits (“Is it safe, kind, fair, or fun?”). And the behavioral correlates reflect fundamental tendencies toward approach and avoidance (historically, the first meaning of attitude: a set or predisposition to respond).

Attitudes not only result in different types of responses—affective, cognitive, and behavioral—but they also result from the same three different types of processes. The major attitude processes that this chapter considers (affective learning, cognitive dissonance, cognitive persuasion, and behavioral response) roughly reflect different relative emphases among these three different types of responses. Readers who relish tidy categories should not push this tripartite distinction too hard; the distinctions often blur and overlap. Nonetheless, it provides a convenient way of talking about various aspects of attitudes.

Attitudes are conceptualized as structures stored in memory (Ferguson & Fukukura, 2012; Judd, Drake, Downing, & Krosnick, 1991). Attitudes differ in their structure, some reflecting the cognitive, affective, and behavioral correlates more than others; the correlates of an attitude show how it covaries (goes along) with the various types of origins and responses (Olson & Zanna, 1993).

Attitude structures can be bipolar, as is typical of controversial issues (Pratkanis, 1989). A bipolar attitude scale would range from pro, through neutral, to anti. For example, most Americans are familiar with arguments on both sides of current debates—immigration, abortion, gun control, death penalty—and could place themselves on that continuum. Other attitudes may be unipolar, representing mainly one's own perspective: matters of taste in music, sports, food, or movies may more often be of the unipolar variety, reflecting less information on the opposing viewpoint. A unipolar attitude scale would range from pro to neutral (liking to indifference). In matters of taste, the negative side of the attitude is less well developed in people's minds (Pratkanis, 1989).

A separate feature of attitude structure is complexity: Attitudes may be complex in that the cognitive component, one's belief system, has multiple dimensions. A simple attitude toward the death penalty, for example, would be “an eye for an eye; murderers must die,” whereas a more complex set of beliefs would consider retribution, incapacitation, deterrence, and just deserts (Carlsmith, Darley, & Robinson, 2002; Darley, Carlsmith, & Robinson, 2000). Complex attitudes often are more moderate ones (Linville, 1982). As the number of purely independent dimensions increases, the more likely it is that one has to consider both pros and cons, becoming more moderate. If the world contains both supporting and contrary information, the more information one obtains—as long as the separate pieces of information are independent of each other—the more moderate one becomes.

Just as attitudes may be cognitively complex, they may be evaluatively complex, in being more or less ambivalent, that is, relatively mixed or unmixed. Ambivalent attitudes tend to be weak (Crano & Prislin, 2006). Such ambivalence can lead to highly variable responses, sometimes on one pole and sometimes on the other. For example, racism can prove ambivalent and volatile (Hass, Katz, Rizzo, Bailey, & Eisenstadt, 1991). Ambivalent white racists see black people as both deviant, eliciting aversive disdain, and disadvantaged, eliciting paternalistic concern. Ambivalence results in more changeable white reactions to blacks who either help or hinder them than to comparable other whites. Similar ambivalence occurs for sexism, in that sexists view some women in benevolent terms (as pure, necessary partners, to be protected) but some in hostile terms (as sexually threatening competitors) (Glick & Fiske, 1996). A later chapter will return to these specific kinds of intergroup ambivalence.

Not only social groups, but also politicians elicit ambivalence; the positive and negative feelings aroused by any given presidential candidate are nearly independent. That is, the correlation between feeling good (hopeful, proud, sympathetic) and feeling bad (afraid, angry, disgusted, uneasy) about a particular political candidate averaged only −.26 (Abelson et al., 1982). For example, many years ago, pride in Edward Kennedy's mystique and care for the common people was offset by disgust at his perceived betrayal of moral standards. In 1998 opinion polls, the American public distinguished sharply between their affective reactions to Bill Clinton's governing (good) and his personal life (bad).

Indeed, attitude objects usually entail separate positive and negative evaluations (Cacioppo & Berntson, 1999). Although people are constrained either to approach or to avoid an attitude object on any given occasion, their reactions probably result from two separate underlying positive and negative affect systems, which combine to produce any given response. Being high on both positive and negative evaluation results in ambivalence.

Attitudes differ too in the consistency between overall evaluation and any of the three cognitive, affective, and behavioral responses; sometimes the heart and mind disagree, as when an affable politician can make people feel warm and fuzzy, even when they disagree completely with his policies. Some attitudes are more internally consistent (e.g., one's feelings match one's beliefs about, e.g., censorship), so they are more likely to polarize, becoming more extreme, the more one thinks about them (Chaiken & Yates, 1985). That is, if all the aspects of the attitude point in the same direction (censorship feels abhorrent, plus one believes that it endangers democracy), then additional thought makes the attitude firmer and more extreme. Conversely, if the heart and mind disagree, additional thought may simply confuse one's attitude (Wilson, Dunn, Kraft, & Lisle, 1989).

Finally, attitudes differ along some other dimensions that will come up later, notably, their importance, strength, and accessibility. Overall, though, the concept of an attitude hinges most centrally on evaluation, with additional consideration of its affective, cognitive, and behavioral correlates, as well as its complexity and internal consistency.

Operational Definitions

Suppose you had to survey employees to find out how sympathetic to corporate merger they are likely to be, a strategy that Daimler-Chrysler's PR firms might well have used to discover and circumvent the potential liabilities of the merger. How might you assess attitudes toward the merger? The most common techniques employ self-reports as operational definitions of attitudes. These comprise the familiar telephone surveys, course evaluations, and customer satisfaction questionnaires. Other, more innovative techniques have developed recently, as a later section will show.

Many self-report techniques start by considering various aspects of the attitude object (corporate merger) and writing items to reflect their salient dimensions. For example: “Corporate mergers fuel the American economy”; “Most corporate mergers have consumers' best interests at heart”; “Most corporate mergers try to protect the rich”; “Corporate mergers generally do not protect their workers” (the last two items, being anti, would be scored in reverse of the first two, being pro). Having written the items, the simplest technique is to use a Likert scale, for example, asking respondents the extent to which they agree or disagree with each item, on a scale from 1 (disagree completely) to 5 (agree completely). People's summed or averaged response to a series of Likert scale items gives the researcher a way to locate each respondent relative to others (Himmelfarb, 1993), for example, as more or less favorable toward corporate mergers. “On a scale from 1 to 10, how much do you like [the merger, the dinner, the President, your school]” is probably the best-known technique for measuring attitudes.

Another common operational self-report definition of attitudes uses semantic differential scales (Osgood, Suci, & Tannenbaum, 1957), again to locate each respondent on a continuum of favorability toward the attitude object. As an example, check one space on each of the following lines to describe your attitude toward corporate mergers: Scores are usually summed from −3 to +3, to provide the respondent's overall attitude, thus locating each person's attitude relative to other respondents' attitudes.

Beautiful	___ : ___ : ___ : ___ : ___ : ___ : ___	Ugly
Bad	___ : ___ : ___ : ___ : ___ : ___ : ___	Good
Pleasant	___ : ___ : ___ : ___ : ___ : ___ : ___	Unpleasant
Dirty	___ : ___ : ___ : ___ : ___ : ___ : ___	Clean
Wise	___ : ___ : ___ : ___ : ___ : ___ : ___	Foolish

Other types of self-report measures, perhaps less familiar, locate the items first, and then the people. In Thurstone scales, judges sort items into, say, nine categories of favorableness. The average rating of each item then determines its scale value. So, for example, an Attitudes toward Blood Donation scale (Breckler & Wiggins, 1989) includes items scaled as low (1 = Blood donation makes me feel ill.), moderate (3 = Blood donation makes me feel bored.), and high (7 = Blood donation makes me feel overjoyed.). Having scaled the items, researchers next administer them to other respondents who then agree or disagree with each item (without knowing the scale values, of course). A respondent's score consists of the average scale value of the item or items endorsed. Louis Thurstone pioneered this technique now named after him. It has the advantage of locating both items and persons along the same continuum from unfavorable to favorable.

The Likert, semantic-differential, and Thurstone scales, supplemented by other self-report measures, all suffer from the same threat that plagues most verbal measures in social psychology: People may distort their responses to accord with socially desirable reactions. Several solutions to the self-report problem have arisen. An antidote called the bogus pipeline (Jones & Sigall, 1971) hooks up laboratory participants to an elaborate but fake psycho-physiological apparatus and leads them to believe it can read their true feelings.

For example, researchers were concerned about whether people answer truthfully when asked about sensitive issues such as their amounts of drinking, smoking, sexual behavior, illicit drug use, and exercise (Tourangeau, Smith, & Rasinski, 1997). Adult volunteers from an urban university community came in for interviews concerning “health questions of a personal nature, such as sexual activity, drug use, and birth control,” to help researchers studying the spread of the AIDS virus. Half the respondents were randomly assigned to a bogus pipeline condition, in which they were told that a physiological recording device could detect inaccurate answers. The bogus pipeline condition made them more honest: It increased their reported frequencies for many of the socially sensitive (perhaps undesirable) behaviors, except for the one socially desirable behavior, exercise, for which reports decreased ( Table 6.1 ). Thus, people were not simply motivated by the apparatus to report higher frequencies of everything (as they might if it made them anxious), but in fact were motivated to be more accurate about socially sensitive behavior that they might otherwise distort. The sheer frequencies should be interpreted with caution, for this is not a representative sample. (Consider the biases in who might volunteer for such a study of personal health.) However, the differences between the two experimental conditions confirm the socially desirable direction of responses under the control condition. Twenty years of research indicate that the bogus pipeline does indeed lead people to respond more truthfully (Roese & Jamieson, 1993). Ironically, because the bogus pipeline depends on this belief, physiological measures of attitudes are trickier than people think (Himmelfarb, 1993).

TABLE 6.1 Self-Reported Socially Sensitive Behaviors, under Bogus Pipeline and Control Conditions

	% Answering Yes
Topic and Item	Control	Bogus Pipeline
Drinking
Drink more than average?	3.4	21.0
Ever drink more than you should?	15.5	33.9
Ever drink and drive?	17.2	30.6
Sexual behavior
Often have oral sex?	32.1	51.7
Illicit drug use
Ever smoke pot?	56.9	71.0
Ever use cocaine?	25.9	43.5
Ever use amphetamines?	19.0	38.7
Ever use other drugs?	19.0	38.7
Other
Do you smoke?	20.7	33.9
Exercise 4 or more times a week?	44.8	22.6

Note: Results appear here only if they were statistically significant or nearly so.

Source: From Tourangeau et al., 1997. Copyright © Winston. Adapted with permission.

Given recent advances in knowledge, sociophysiological measures can be useful (Blascovich & Mendes, 2010): One aspect of EEG (electroencephalogram) brain waves (namely, late positive potentials) represents an evaluative categorization (essentially, a primitive attitude measure) in response to specific positive and negative stimuli (in this case, personality traits) (Cacioppo, Crites, Gardner, & Berntson, 1994). Similarly, fMRI (functional magnetic resonance brain imaging) methods are beginning to allow social psychologists to locate approach and avoidance reactions in particular areas of the brain (e.g., Ochsner & Lieberman, 2001). For example, distinct brain areas respond to the positive–negative valence (right insula) and emotional intensity (amygdala) of an attitude (Cunningham, Raye, & Johnson, 2004). While an intriguing solution to the self-report problem, neuroscientific measures are still cumbersome, expensive, and beyond the current expertise of many social psychologists, not to mention persuasion practitioners in the larger world. What's more, they are controversial because they tend to be disproportionately believable. For example, legal scholars worry that expert witnesses who present brain-imaging data will come across as unassailable (Tancredi & Brodie, 2007), raising ethical issues (Farah, 2012). But others argue that neuroimages have less impact than feared (Roskies, Schweitzer, & Saks, 2013). Neuroscience increasingly contributes to measuring the affective underpinnings of attitudes, as this chapter will show.

Another antidote to the self-report problem comes from an unobtrusive measure, namely, any measure that does not impinge on respondents, in that they do not know it is being taken. For example, nonverbal behavior is sometimes an unobtrusive measure of attitudes. Interracial attitudes can be operationalized via nonverbal indicators of immediacy that have included physical distance, forward lean, eye contact, and shoulder orientation (i.e., how directly one person faces the other), as well as length of interview (Word, Zanna, & Cooper, 1974).

Other kinds of unobtrusive measures are more cognitive. For example, cognitive activation can unobtrusively measure racial attitudes (Fazio & Olson, 2003; see a later section and also the prejudice chapter). To the extent that one negative category is primed or activated (mentally available;  Chapter 4 ), another, immediately following negative category is recognized more quickly. That is, a person first primed with cancer responds more quickly to the next word when it is crime (negative) than candy (positive). In one study (Fazio, Jackson, Dunton, & Williams, 1995), color photographs of students from various racial groups served as positive and negative primes, to the extent a respondent was prejudiced. Students first viewed a photograph and then a positive or negative adjective. They more quickly responded good or bad when the adjective's evaluation matched their race-based evaluation of the photograph. This facilitation or speed-up of evaluatively matching responses (own racial group—good, other racial group—bad) indicated their negative judgments of other-group races and correlated with various racial attitude scales. This measure clearly circumvents the self-report problem. The Implicit Associations Test relies on a similar method (Greenwald, McGhee, & Schwartz, 1998; see later sections of this chapter and the prejudice chapter).

Finally, some novel and entertaining unobtrusive measures focus on group-level data. At the Museum of Science and Industry in Chicago, researchers once measured interest in various exhibits by how often the museum had to replace the floor tiles in those locations (the chick-hatching exhibit was tops). They could even estimate the relative ages of people interested in various exhibits by the height of the nose-prints on the glass cases. These and other oddball examples of unobtrusive measures (collected by Webb, Campbell, Schwartz, Sechrest, & Grove, 1981) typically do not focus on individual respondents, but rather on group-level data (e.g., all the museum visitors over a certain time period), so social psychologists, who are typically interested in individual behavior, rarely use them. But they remind us that the measurement of attitudes is limited only by the creativity of the researcher.

Core Social Motives as Functions of Attitudes

Early attitude theorists (e.g., Katz, 1960; Smith, 1947) pointed out that attitudes can serve various functions for people; these functions correspond to the core social motives that run through all of social psychology in one form or another. Attitude research focuses on two, which here are called understanding and belonging ( Table 6.2 ).

TABLE 6.2 Attitude Functions and Core Social Motives

Attitude Function	Core Social Motive
Object appraisal (+/−)	Understanding (approach/avoid)
Utilitarian (specific goal helped or hindered)
Value Expressive	Belonging
Public: Social adjustive
Private: Pure value-expression

OBJECT APPRAISAL: UNDERSTANDING

The clearest function of attitudes is object appraisal, which fits our core social motive of understanding: Attitudes categorize entities, so the person can decide which to approach or avoid. The object-appraisal function may hold universally, for all attitudes and all people (Eagly & Chaiken, 1998). Object appraisal generally facilitates people's goals in the world because people first form categories and then instrumentally assess whether the category helps or hinders the goal. Object appraisal is a primitive form of categorization and as such overlaps with a basic knowledge function, which furthers rapid understanding of one's world.

The automatic accessibility of one's attitude, that is, having an evaluation immediately upon encountering an object, fits the broad view of object-appraisal functions (Fazio, 1989). In keeping with a social adaptation view, the general object-appraisal function is good for your health. In other words, knowing what one feels, and therefore not having constantly to reconstruct or to struggle with one's preferences, relieves stress. Being in limbo, wallowing in indecision, worsens stress. The distraction, vigilance, worry, and preoccupation evidently all take a toll.

In one study (Fazio & Powell, 1997), researchers found that first-year college students varied in the extent to which they had accessible (rapid) attitudes toward academics—that is, some instantly knew their attitudes toward various classes, majors, schedules, and study strategies; others did not. The decisive students, who had their minds made up with more accessible attitudes, were buffered in dealing with life stresses. For example, consider initially healthy first-year students (depicted by the two solid lines in  Figure 6.1 ). If major adversity (e.g., family problems) came into their lives, as reflected by their reported stress scores (horizontal axis of the figure), everyone was more likely to become physically ill (vertical axis). However, the students with accessible attitudes (circle endpoints in the figure) showed less impact of life stresses on illness, compared with those with inaccessible attitudes (triangle endpoints). That is, the illnesses of students with accessible attitudes did not increase as steeply with their life stress. In short, accessible attitudes diminished the stressors' negative impact, resulting in fewer doctor visits, days lost to illness, and physical symptoms, compared with students with less accessible attitudes, who presumably had to agonize over academics as well as cope with other life stresses. In effect, knowing what you want saves the added burden of indecision, if you are stressed but otherwise healthy.

Figure 6.1  First-Year Students' Physical Illness, as a Function of Initial Illness, Life Stress, and Attitude Accessibility

Source: From Fazio & Powell, 1997. Copyright © Blackwell. Adapted with permission.

In contrast, initially unhealthy students with accessible attitudes (in the figure, dotted lines with circle endpoints) benefited only at low levels of life stress; presumably, high levels of stress, coupled with ill health, either put them over the top or reflected their inability to cope effectively in the first place. Thus, apart from overwhelming health problems and life stress, the main point is that rapid object appraisal can serve a real function in people's lives.

A more delimited kind of appraisal function is the more narrowly utilitarian function, using attitudes to further one's specific self-interests. Attitudes can facilitate one's access to money or other resources, as one example (Eagly & Chaiken, 1998). Consider politicians who change political parties to get elected more easily in the other party; consider lawyers who adopt a positive attitude toward their clients—whether justified or not—in order to plead more effectively for them; consider the employee who adopts a positive attitude toward a corporate merger, hoping to advance in the company faster than a recalcitrant employee. Actually, people overestimate the effects of utilitarian self-interest on other people's attitudes (Miller & Ratner, 1998), so maybe it is a less important function than one might think. In any case, the utilitarian function has a narrower purview than the pure object-appraisal function, which entails understanding in a broader context ( Table 6.2 ). Although object appraisal may be universal, people differ in their need to evaluate, that is, the degree to which they report having attitudes and thinking evaluative thoughts (Jarvis & Petty, 1996). People high on need to evaluate respond in both more positive and more negative terms to paintings, politics, and life events. One might say they have an opinion about everything. (This need to express one's evaluations differs from a need to think, so the omnipresent opinions may or may not be deep!) As with all the core social motives, object appraisal drives social responses for everyone, but perhaps for some even more than for others.

VALUE EXPRESSION: BELONGING

In contrast to understanding for self-relevance—the main message of object appraisal—value-expressive functions align with important standards or social approval (Olson & Zanna, 1993); value-expressive attitudes represent one's identity, either publicly or privately (Shavitt, 1990). Thus, value expression creates a sense of belonging to desirable social groups, by seeming to share their attitudes and underlying values. The public side of this has been termed a social adjustive function (“I want to be part of that group who defends civil liberties”), and the private side has been termed pure value expression (“I'm the kind of person who believes in civil liberties”).

The contrast between object appraisal and value expression as a functional basis of attitudes matters in a practical way. For example, attitudes toward people with AIDS can serve an object-appraisal function (e.g., fear of contamination), but they can also serve a value-expressive function (e.g., antipathy toward homosexuals). Educating employees about the improbability of AIDS transmission in the workplace—addressing the knowledge or object-appraisal function—had no impact on people with negative attitudes toward homosexuals, for whom AIDS symbolizes alleged homosexual promiscuity and moral decadence, serving a value-expressive function (Pryor, Reeder, & McManus, 1991). Presumably, attitude change would require addressing the values underlying attitudes toward people with AIDS, emphasizing not only knowledge and object appraisal but also the value-expressive function of the attitudes.

WHICH FUNCTION WHEN?

Some people, some situations, and some attitude objects elicit the more practical function of object appraisal, and some elicit the more social value-expressive function (Tesser & Shaffer, 1990).

Start with people: Some people emphasize attitudes' public and others their private aspects, even within the value-expressive function (e.g., social adjustive vs. pure value-expression). As noted in the self chapter, and as a later section elaborates, high self-monitors (who are more sensitive to the public, social environment) respond differently than low self-monitors (who are more sensitive to their private, internal values). Consequently, high self-monitors should focus more on public images in advertising, responding to the soft sell, while low self-monitors should focus more on alleged quality, responding to the hard sell. Researchers (Snyder & DeBono, 1989) tested these hypotheses by creating ads for products as varied as coffee, cars, whiskey, shampoo, and cigarettes, and they varied only the ad copy to create a soft sell (image) or hard sell (information). One ad showed a bottle of Canadian Club whiskey sitting on a set of house blueprints, with copy featuring either image (“You're not just moving in, you're moving up.”) or information (“When it comes to great taste, everyone draws the same conclusion.”). Indeed, high (compared with low) self-monitors found the image-oriented ads better, more appealing, and more effective, plus they were willing to pay more money for the product and were more likely to try it. Apparently, the attitudes of high self-monitors serve a social-adjustive (public belonging) function; they care about how their attitudes help them to function with group members they value. In contrast, the attitudes of low self-monitors serve an internal value-expressive (private belonging) function, although low self-monitors may also be likely to emphasize utilitarian and knowledge (pure object-appraisal) functions.

Situations also emphasize different attitude functions. In one study, some ads urged donations to cancer research by emphasizing value-expressive (“help people to live”) or utilitarian object-appraisal (“protect your future”) functions. Values about altruism and helpfulness predicted attitudes toward donation in the value-expressive condition only (Maio & Olson, 1995). In contrast, pressures toward snap judgments would engage the object-appraisal function (Jamieson & Zanna, 1989).

Finally, some attitude objects tend to elicit certain functions more than others. For example (Prentice, 1987; Shavitt, 1990), attitudes toward family heirlooms and bibles serve symbolic (belonging, value-expressive) functions, whereas attitudes toward televisions and bicycles primarily serve utilitarian functions.

Summary of Definitions and Motives

Attitudes are primarily evaluative responses to entities, with affective, cognitive, and behavioral correlates. Social psychologists usually measure attitudes by verbal self-reports on numerical scales. All attitudes serve the object-appraisal function of understanding: categorizing entities with favor or disfavor, tendencies to approach or avoid. Some attitudes fill this object-appraisal function more narrowly as utilitarian attitudes serving particular self-interested goals. Besides object appraisal, the other major function is value expression. In the public sense, this is also social adjustive, and in the private sense, this is also pure value expression, but both promote identity with groups and standards and therefore a sense of belonging.

HOW ATTITUDES FORM VIA AFFECT FIRST: UNDERSTANDING WHAT TO APPROACH OR AVOID

The average citizen has shockingly few real political attitudes, defined as those that display much knowledge and stability. Ideology is beyond the reach of what most Americans care to manage (Kinder, 1998). The prevalence of nonattitudes—insubstantial political fads and whims, here today and gone tomorrow—would seem to threaten the basic principles of democracy, which presumes a thinking public (Krosnick & Visser, 2010). Yet, as Kinder puts it felicitously, if unkindly, American citizens are awash in ignorance, show intolerance galore, and are unsophisticated in the extreme. For example, “in late 1995, more than twice as many Americans believed, stupendously incorrectly, that the federal government spends more money on foreign aid than on Medicare” (Kinder, 1998, p. 785). Nevertheless, when the pollster calls, people willingly express opinions that we spend too much public money bailing out foreign countries and too little taking care of our own elderly. Whatever their proclivities, having the underlying facts would make their opinions more compelling than when they are innocent of the facts. People respond out of “wretched ignorance,” with bias against opinions and groups that differ from their own, and with a preference for easy answers. This shocking accusation becomes commonplace, however, when one examines how attitudes form. The opinion (knowledge) component does not figure greatly in most attitude acquisition, as it happens.

Instead, affective processes predominate. People want to understand what to approach and what to avoid (Cacioppo & Berntson, 1999). Arguably, knowing a bunch of obscure facts matters less to democratic politics than does having a consistent attitudinal evaluation (Krosnick & Visser, 2010). If attitudes are nothing if not evaluative, according to the usual definition, then it makes sense that the most basic forms of attitude acquisition emphasize affective processes, not cognitive content. Attitudes have both societal and biological sources (Albarracín & Vargas, 2010), which may respectively contribute to the more cognitive and more affective aspects. This section addresses some affectively oriented forms: learning theories, primary emotional appraisals, and mere exposure, all unreasoned attitudes. Then it addresses relatively unconscious evaluative associations, implicit attitudes. It closes with a more focused discussion of emotions and attitudes.

Learning Theories

One of the most basic and affective attitude formation processes, conditioning, does not surprise laypeople (Snell, Gibbs, & Varey, 1995); people's everyday theories of attitude formation fit conditioning, perhaps because elementary principles of learning have filtered into society. And no wonder. Starting with rats, pigeons, dogs, and the occasional cat, researchers mapped the outlines of fundamental learning processes during half a century. Although the origins of this research lay in animal models, the generalization to humans proved fruitful. The next two brief sections summarize some basic principles of animal learning theories, in order to apply them to people. The power of the situation (learning) shapes affective and behavioral preferences.

CLASSICAL CONDITIONING

Introductory psychology classes teach the principles of classical conditioning, as famously uncovered by Ivan Petrovich Pavlov and his nameless dog. Repetition and association were the combination that unlocked classical conditioning. The dog initially, spontaneously, reflexively salivates upon encountering food powder. Salivation is an unconditioned response (UCR) to food, which is the unconditioned stimulus (UCS). To this preexisting UCS-UCR pairing (food-salivation) is added the to-be conditioned stimulus (CS), in this case a metronome beat. As the metronome beat is repeatedly paired with food, it too comes to elicit the conditioned response (CR, here salivation), until it can do so independently. The UCS-UCR pairing becomes a CS-CR pairing.

An analogous process operates in the classical conditioning of attitudes (Staats & Staats, 1958). People naturally dislike (UCR) electric shock (UCS), and when shock is repeatedly paired with certain words (CS), the words take on a negative association (CR). One could imagine a child responding, reflexively, negatively, to the unconditioned stimulus of a parent's nonverbal cues of fear or anger. When those are paired repeatedly with a particular political party or various ethnic groups, the child may develop the same negative associations to those groups. Classical conditioning accounts for a range of attitudes (Walther, Nagengast, & Trasselli, 2005).

Classical conditioning works best when prior knowledge is low (Cacioppo, Marshall-Goodell, Tassinary, & Petty, 1992). Neutral words (e.g., player) condition less effectively than equally neutral nonwords (e.g., trames). When researchers pair words with shocks, they are disliked, but not as much as nonwords paired with shocks ( Figure 6.2 ). This result indicates that classical conditioning works best with unfamiliar stimuli, which is consistent with viewing conditioning as a form of initial, affect-based attitude acquisition. Classical conditioning results in an attitude represented by feelings, predicting impulsive and nonrational (e.g., affectively based) behaviors.

Figure 6.2  Classical Conditioning of Attitudes toward Words and Nonwords

Source: From Cacioppo et al., 1992. Copyright © Elsevier. Adapted with permission.

Other research (Kim, Lim, & Bhargava, 1998) supports this interpretation. The direct transfer of affect in classical conditioning (i.e., the positive affect associated with a close-up of a kitten), over repeated associations, can create attitudes toward irrelevant consumer products such as “Brand L Pizza House,” in the absence of any beliefs about the product. Thought-listing in response to the kitten photos included “cute, soft, playful, and curious,” not attributes related to pizza. Nonetheless, the kitten-related positive feelings transfer to the pizza. Moreover, affect can outweigh beliefs in attitude formation, as when the photo and the pizza house were paired ten times.

Classical conditioning works even for people as stimuli: Pairing a neutral person with someone liked or disliked spreads affect not only to that person but also to another person associated with that second person (Walther, 2002); be careful who your friends are! Classical conditioning elicits pure affect, and it fits the object-appraisal function of attitudes, namely the motive to understand which objects are “good for me” and which “bad for me.” Conditioning may especially underlie attitudes that are less than fully conscious (Olson & Fazio, 2001).

INSTRUMENTAL CONDITIONING

Instrumental (also known as operant) conditioning works on this simple principle: Animals emit various random behaviors, some reinforced (rewarded or punished) by the environment, which in turn changes the frequency of the behavior. In B. F. Skinner's laboratory, pigeons pecked around their cages, until randomly pecking a key that released a food pellet; that reward encouraged the pigeon to repeat the behavior. Instrumental conditioning is aptly named because the consequence of the behavior becomes its (instrumental) cause. The behavior is repeated in order to obtain rewards or avoid punishments, collectively known as reinforcements.

Instrumental conditioning of attitudes is illustrated by studies in which telephone interviewers say “good” every time the respondent happens to emit a response in the desired direction. For example (Insko, 1965), students at the University of Hawaii, subtly conditioned in this way, developed positive or negative attitudes toward creating a springtime Aloha Week, attitudes that held up as much as a week later.

Among researchers, an intense debate raged for years over whether or not attitudes are conditioned only when people become aware of the contingency between the attitudes they express and the verbal reinforcement. If awareness is necessary, perhaps people's attitudes change only because they are trying to please the experimenter or conform to social norms. If so, conditioning might be a trivial methodological artifact, not the presumed affective learning process. In the end, although beyond the scope of this text, awareness probably is not required for the instrumental conditioning of attitudes (Eagly & Chaiken, 1993).

SOCIAL LEARNING

Social learning theories introduce uniquely interpersonal processes into the conditioning equations. Two kinds of social learning theories provide respective conceptual parallels to the classical and instrumental conditioning of attitudes. A subsequent chapter elaborates, but the initial comparison is useful now. Classical conditioning and instrumental conditioning each have interpersonal analogues in social learning.

A process termed modelling essentially describes learning by imitation of another person, a process that roughly parallels classical conditioning, thus: An admired other (older sibling, parent, popular peer) repeatedly endorses a presidential candidate (or any other opinion), and the positive feelings (UCR) associated with the admired other (UCS) become associated as a conditioned positive response (CR) to the opinion (favoring the presidential candidate) (CS). Although other interpretations are possible (Lott & Lott, 1985), modeling or imitation constitutes a relatively simple form of social learning.

Another process, termed vicarious conditioning or observational learning, roughly parallels instrumental conditioning. Watching another person receive rewards and punishments allows observers to shape their own behavior according to those reinforcements without having to experience them directly (Berger & Lambert, 1968). How many teenagers have learned the cool attitudes by watching peers reward or punish other teens for endorsing certain music, styles, or convictions? How many children have watched another child in trouble with the teacher and learned to avoid the same provocative behavior and its inevitable consequences? Later chapters on helping and on aggression take up social learning in more detail, but for now, social learning illustrates some processes of attitude formation, emphasizing affect. Again, it relates most closely to the object-appraisal function of attitudes, understanding what features of the environment to approach and avoid.

Emotional Appraisal

A broader view of object appraisal has long appeared in emotion theories, and it fits with newer attitude approaches that incorporate complex emotions into the basic approach-avoidance view of attitudes. Emotion theories of attitudes distinguish immediate primary appraisals (good for me, bad for me) from subsequent secondary appraisals, a more complex reaction implicating causality, responsibility, certainty, and timing (Ellsworth & Scherer, 2003). Primary appraisals resemble the approach-avoid, positive-negative object appraisals we have seen so far, but emphasizing their speed. More specific secondary emotional appraisals result from specific events (Schwarz & Clore, 2007). For example, anger implies an illegitimate negative outcome controlled by another person, whereas guilt implies a negative outcome controlled by the self. By going beyond simple evaluative valence, specific emotions thus distinctively influence judgment and behavior, in predictable ways, based on the underlying appraisals (Lerner & Keltner, 2000).

Specific emotional appraisals shape attitudes in ways that go beyond valence. For example, inducing anger causes automatic negative evaluations of an outgroup, whereas inducing sadness or neutrality does not, presumably because they are normally less relevant to intergroup relations (DeSteno, Dasgupta, Bartlett, & Cajdric, 2004). Similarly, anger and sadness each elicit distinct attitudes regarding policy preferences ranging from social welfare to terrorism (Small & Lerner, 2008; Small, Lerner, & Fischhoff, 2006). Political candidates who exploit fear, anger, or hope know their distinct implications for voting behavior.

Mere Exposure

Most political campaign tactics do not afford much thought, and the minimalist campaign outlet consists of a billboard simply stating the candidate's name and the office sought. The more the billboards, the more voters like the candidate, and conditioning is not required; no positive associations or rewards matter. Mere exposure, the sheer frequency of encountering an initially neutral or positive stimulus, enhances evaluations of it, a now-classic phenomenon discovered by Robert Zajonc (1968). Indeed, frequency of exposure predicts attitudes toward Chinese ideographs and yearbook photographs, as well as names; it is a small leap from here to bumper stickers as a means of establishing favorable attitudes toward a candidate. In a 20-year review (Bornstein, 1989), mere exposure studies show a moderate correlation of about .26 between frequency of exposure and liking, but the effect is variable. Stronger exposure effects hold for adults viewing photographs, meaningful words, and polygons, when they are presented 10–20 times, under brief exposure (less than a second), and later rated for liking, after a considerable delay. Fortunately for advertisers, the effects are even stronger in natural settings than in the laboratory.

Most impressive is evidence that the mere exposure effect does not depend on the exposure itself being conscious (Murphy et al., 1995), nor does it even depend on later recognizing the frequent stimulus as familiar (Moreland & Zajonc, 1977). In the world, more common than unconscious (subliminal) advertising is the technique of frequent and broad exposure of a product. Similarly, the political publicist maximizes airtime, in keeping with the adage, “no such thing as bad publicity”; whatever gets the name out, it helps (although the mere exposure effect suggests the context has to be neutral or positive).

One mechanism appears to be perceptual fluency (ease of processing), which produces positive affect, as indexed by micro-muscular movements in the smile muscles (zygotmaticus major; Winkielman & Cacioppo, 2001). More generally, processing fluency—whether based on aesthetic principles (e.g., figural goodness, figure-ground contrast, stimulus repetition, symmetry, prototypicality) or priming—increases pleasure (Reber, Schwarz, & Winkielman, 2004). Consumers often base choices on products that “feel right,” because they are encoded or remembered easily (Ariely & Norton, 2009). For example, on the New York Stock Exchange, stocks with fluent (pronounceable) ticker codes outperform those with unpronounceable ticker codes (Alter & Oppenheimer, 2006). Again, though, attitudes formed this way have no cognitive content, only affect.

Automatic and Implicit Attitudes

The rapidity of forming good-bad, approach-avoid appraisals means that these processes often operate with minimal or no conscious awareness. Prejudice researchers particularly need to understand people's hidden biases (see that chapter), but implicit and automatic attitudes apply to all kinds of object appraisals (Fazio & Olson, 2003). We have already discussed attitude accessibility and facilitation of responding to attitude objects, following matched, evaluative-consistent primes.

Another well-known theory concerns implicit associations, people's simultaneous links between the attitude object and positive or negative words (Fazio & Olson, 2003). This phenomenon concerns implicit attitudes because it displays traces of past experiences (i.e., associations) that are not accurately represented in conscious awareness. One of the most convincing demonstrations is the Implicit Associations Test (IAT; Dasgupta, McGhee, Greenwald, & Banaji, 2000; Greenwald et al., 1998). At this writing, the IAT is available on the Internet (put the name of the test into any standard search engine), so one can take the IAT online; for example, one may associate presidential candidates with good and bad concepts, showing one's subtle evaluations of them. The IAT shows how rapidly and easily—implicitly—people do this.

The IAT has invited more controversy than usual for an attitude theory, perhaps because of its use in measuring racial attitudes (see this and the prejudice chapter)—or perhaps more generally, associations that people prefer not to report on a more deliberate, explicit basis. Many observers ask whether the “real” attitude is the implicit attitude (e.g., the faster association between career and male, versus family and female) or the explicit endorsement (e.g., reporting egalitarian attitudes). Consequently, the IAT has sparked considerable efforts to validate the meaning of the measure (Greenwald, Nosek, & Banaji, 2003; Greenwald, Poehlman, Uhlmann, & Banaji, 2009; Nosek, Greenwald, & Banaji, 2005; Olson & Fazio, 2003). At this point, meta-analyses (more than 100 studies on more than 10,000 participants) indicate its predictive validity for a range of behavior, physiology, and judgment (Greenwald et al., 2009). The IAT is especially effective in socially sensitive domains, adding incremental validity to explicit self-reports in delicate areas such as prejudice. Overall, implicit attitudes express affect-laden associations that people may or may not endorse, that come from either cultural or personal experience; in contrast, explicit attitudes express propositions that people regard as true, involving more overt, logic-oriented (but not necessarily objective or accurate) reasoning (Gawronski & Bodenhausen, 2006).

The Importance of Affect and Emotion

Why do theories of attitude formation tend to emphasize affect and to neglect cognition? Affect is the foundation of attitudes. As we will see, most theories of attitude change, as opposed to attitude formation, focus at least equally, if not more heavily, on cognition. Those theories of attitude change (persuasion) assume an existing attitude and mostly cognitive processes to change it. In contrast, the theories of attitude formation assume no prior attitude. The relative emphasis on affect in theories of attitude formation reflects the primacy of affect (Zajonc, 1980); that is, affective reactions, relative to cognitive ones, are more immediate, involving, inescapable, irrevocable, and compelling. Thus, affect enters in at the first stages that make an attitude an attitude (a reaction with a strong affective-evaluative component). Recall that the chapter opening defined an attitude as, at a minimum, an evaluative (affective) reaction. Moreover, people react faster in reporting how they feel than how they think about various attitude objects, such as brand names and countries (Verplanken, Hofstee, & Janssen, 1998). Affect is primary in many respects, although people differ in their reliance on heart and mind as bases for their attitudes (Haddock & Zanna, 1998). Subsequent work confirms that the affect is primary mainly for attitudes with a heavy evaluative basis (Giner-Sorolla, 2004).

The Zajonc position on the primacy of affect caused quite a controversy in the field (Fiske & Taylor, 2013, Chapter 14). For example, some (e.g., Epstein, 1984; Lazarus, 1982) argued that primitive forms of nonconscious cognition, object appraisals, have to precede emotion. Some forms of cognition can be just as rapid as affective responses and not necessarily more rational (Holyoak & Gordon, 1984). Zajonc (1984) countered that cognition must be distinguished from initial perception; his point was to contrast more intellectual from more affective forms of understanding. The current consensus is that affect and cognition are deeply interdependent (Storbeck & Clore, 2007), in attitudes as elsewhere, although cognitive approaches tended to dominate the field in certain eras, notably during the 1950s communication-and-persuasion emphasis and during the 1980s dual-process persuasion models, both covered in later sections.

The importance of emotion in attitudes has become even more evident with new theories and methods for understanding social emotions (Keltner & Lerner, 2010). For example, attitudes are embodied or represented by consistent physical enactments (recall embodied cognition from Chapter 3). People express approach (usually positive) evaluations more rapidly by pulling a lever toward themselves and avoidance (negative) evaluations more rapidly by pushing it away (Centerbar & Clore, 2006). Nodding while listening to a strong persuasive communication facilitates agreement, whereas shaking one's head facilitates disagreement (Briñol & Petty, 2003). Beyond physical representations of attitudes, affective neuroscience implicates the amygdala and right insula, as noted earlier, plus reward areas such as orbital frontal cortex (O'Doherty, Kringelbach, Rolls, Hornak, & Andrews, 2001). Earlier, we noted the involvement of the amygdala in affect intensity; in this vein, people viewing politicians showed amygdala activation that correlated with strength of emotion (Knutson, Wood, Spampinato, & Grafman, 2006). Whatever way affect emerges, the core of attitudes emerges in such evaluations and more complex emotions.

Logically, attitude formation could employ either cognitive or affective processes to form an initial attitude. For example, one study (Edwards, 1990) compared affect-based and cognition-based attitudes, to see whether they most easily change by persuasion methods that match, respectively, being affective or cognitive. Participants formed attitudes toward Chinese ideographs, either affectively first or cognitively first. In the affective formation, a smiling or frowning face preceded the slide for the ideograph; the face appeared so rapidly (10 milliseconds) that it was outside awareness but could still influence affective reactions to the ideograph that followed. In the cognitive formation, an informative paragraph described the calligraphy favorably or unfavorably. Participants then rated the ideographs, indicating that positive and negative attitudes had formed. Then the sequence repeated, in order to try to change the attitude. Change was most effective when it matched the basis of persuasion: Subliminal faces changed attitudes that had formed using subliminal faces, and information-based paragraphs changed attitudes that had formed by information-based paragraphs ( Figure 6.3 ). Thus, although evaluation is essential, and emotions are deeply involved, some attitudes are even more affective and some relatively more cognitive than others.

Figure 6.3  Match between Cognitive or Affective Basis for Attitude and Most Effective Form of Persuasion

Source: From Edwards, 1990. Copyright © American Psychological Association. Adapted with permission.

Can People Control Affectively Formed Attitudes? Prejudice as a Critical Case

Classical, instrumental, and social learning, emotional appraisals, mere exposure, and implicit associations—as well as the more emphatically emotion-oriented approaches—together raise the question of how much we control our attitudes. As noted in the social cognition chapter, more of our reactions are automatic than we like to think (Bargh & Chartrand, 1999), and where affect is concerned, the question of control becomes even more dicey. The limits of our control are moderated, however, by individual differences, situational motivators, and context. The attitude domain where this has been most aptly demonstrated is prejudice, precisely because most people would prefer to control their prejudices. A later chapter addresses these biases in detail, but for now they illustrate the automaticity and control of affect-laden attitudes.

INDIVIDUAL DIFFERENCES IN CONTROL

People differ in their motivation to control their own expressed attitudes, and two types of cognitive processes reflect this motivation. Researchers typically compare relatively automatic, subtle, implicit, spontaneous reactions with relatively controlled, blatant, explicit, deliberate reactions. The more automatic reactions allow less capacity for control, and the relatively deliberate responses allow more capacity for control. We would expect people to be able to control the more deliberate responses, if they are motivated, because the capacity is present. But to what extent do they control the more spontaneous responses, where capacity is limited, even if they are motivated?

People experience a need to control their responses when they violate their own standards for acceptable behavior. Everyone sometimes experiences discrepancies between the way they feel they should behave and the way they do behave, as the self chapter indicated. Control over one's prejudice exemplifies this rule (Devine, Plant, & Buswell, 2000; Monteith & Mark, 2005). For example, people with large should-would discrepancies might laugh at a racist or sexist joke, failing to monitor their own behavior, especially if they are distracted (Monteith & Voils, 1998). Besides should-would discrepancies, researchers more often measure individual differences in prejudice and split people into low- and high-prejudice groups on that basis. Both high- and low-prejudice people can have should-would discrepancies, which potentially motivate them to behave better, but the dynamics differ for high- and low-prejudice people.

Low-prejudice people experience a discrepancy when cultural stereotypes surface automatically in their own responses, which then conflict with their own unbiased beliefs (Devine, 1989). When they do transgress their own internalized standards, low-prejudice participants slow down and think about the discrepancy, and they feel compunction, guilt, and self-criticism. As a result of discrepancy, they can effectively inhibit their spontaneous affective reactions, such as laughing at biased jokes (Devine, Monteith, Zuwerink, & Elliot, 1991; Monteith, 1993; Monteith, Devine, & Zuwerink, 1993). Low-prejudice participants can inhibit stereotype activation under both automatic and controlled conditions (Kawakami, Dion, & Dovidio, 1998; Lepore & Brown, 1997). Thus, they make the motivated effort and indeed have some success at controlling their prejudices, even automatic ones. Compunction plays an important role in overriding spontaneous affect-laden attitudinal responses.

In contrast, more prejudiced people readily endorse and activate cultural stereotypes (Kawakami et al., 1998). Highly prejudiced people do have standards for what constitutes unacceptably biased behavior; they have their own moral obligations about being tolerant, particularly when they focus on equality of opportunity (Monteith & Walters, 1998). People who are relatively prejudiced sometimes feel guilty about their transgressions, even though their standards are lower, relative to less prejudiced people's high internal standards (Monteith, 1996). More prejudiced people sometimes feel uncomfortable and angry about being constrained by other people's higher standards (Devine et al., 1991; Monteith et al. 1993). (A better description of resentment around political correctness I cannot imagine.) The bottom line is that highly prejudiced people make less effort to inhibit their automatic attitudes and may be more likely to blame others when they themselves go astray.

Besides measuring people's prejudice directly, researchers sometimes use known groups, that is, categories of people expected a priori to differ on a given attitude (e.g., prejudice). For example, black students differ on average from nonblack students in antiblack prejudice, both the implicit and explicit kinds (Rudman, Ashmore, & Gary, 2001). In another known-groups variant, students who had just taken a seminar on prejudice and conflict might well be expected to differ on average from those enrolled in a large lecture course or a research methods course. Although they did not differ at the beginning of the course, by the end of the course, their scores on both implicit and explicit measures of both stereotyping and prejudice improved (Rudman et al., 2001). Over the longer term, groups of whites who report having had both unprejudiced parents and some childhood experiences with blacks show greater explicit concern with avoiding discriminatory actions, although their automatic reactions differ only if their positive interracial experiences have been recent (Towles-Schwen & Fazio, 2001).

Known groups also differ in their attitudinal dynamics. Dominant group members (i.e., whites) worry about how they themselves are evaluated in intergroup interactions (Vorauer, Main, & O'Connell, 1998). Some whites are more interracially anxious than others (Britt, Boniecki, Vescio, Biernat, & Brown, 1996). Low-prejudice white people have a more negative meta-stereotypes (beliefs about what other groups think of their group), but high-prejudice people feel more personally stereotyped by the outgroup (“Why do those black people always assume I am prejudiced just because I am white?”). Meta-stereotypes activate when people are concerned about being socially evaluated by the outgroup member, and it matters to them (Vorauer, Hunter, Main, & Roy, 2000). Whites who are concerned about appearing prejudiced are more anxious and enjoy an interracial interaction less, but ironically their black partners enjoy it more, suggesting that the white person's efforts to inhibit prejudice are costly but effective (Shelton, 2003). Indeed, managing interracial contact can be distracting for prejudiced whites (Richeson & Shelton, 2003).

Criteria less obvious than race also can guide the selection of known groups to assess attitudes. For example, one set of basketball fans with a reputation for expressing enthusiastic and overt outgroup hostility differed from another group of fans known for its more restrained style (Franco & Maass, 1996). The more aggressive fans showed more prejudice and discrimination on explicit measures (reward allocation and trait attributions) but no difference on more subtle measures (biased language), suggesting that the relatively automatic attitude is indeed more implicit, harder for people to control. As another example, older people might be less able to inhibit their reactions than younger people (von Hippel, Silver, & Lynch, 2000). Although they report feeling more motivated to control their prejudice, the older people's weakened inhibitory capacity made them respond in more prejudiced and stereotyped ways than they wanted, even on subtle (but not automatic) indicators of stereotyping. Both capacity and motivation are necessary, and the older people may lack the capacity, though not the motivation.

Besides the work splitting people on high- and low-prejudiced attitudes, using known groups, other work examines correlations between people's implicit and explicit (roughly, automatic and controlled) attitudes. Here's the logic: If the more spontaneous, implicit, automatic attitudes are indeed not controllable, then they should diverge from the controlled ones for people motivated and able to exercise that control. Presumably, the automatic attitudes would be more prejudiced and the controlled ones more or less prejudiced, so their correlation would be low. If the relatively automatic ones instead are controllable, then the two should correlate. Both would depend on the person's motivation and ability to control prejudice. Sometimes the two kinds of attitudes correlate nicely, suggesting the possibility of control (e.g., Rudman et al., 2001; Wittenbrink, Judd, & Park, 1997), and sometimes they do not, suggesting the difficulty of control (e.g., Brauer, Wasel, & Niedenthal, 2000; Karpinski & Hilton, 2001; von Hippel, Sekaquaptewa, & Vargas, 1995, 1997).

Motivation to control prejudice also moderates people's ability to control their automatic responses (Payne, 2001). In a direct parallel to the circumstances that can cause police mistakenly to shoot unarmed black men, white people misidentify a hand tool as a gun more often when it is associated with a black male face than when it is associated with a white male face. They are most likely to make this (potentially tragic) mistake when they are operating on automatic, with little time to consider their response. For those not at all motivated to control bias, their automatic bias directly relates to their explicit racism, again because the failure to control leaves the two types of response free to be the same. For those highly motivated to control bias, their racism scores are inversely related to their explicit racism (i.e., the higher their racism, the more they have to control the automatic bias). These studies again show that capacity (automatic versus controlled levels of time to respond) and motivation (effort to control bias; lower racism scores) both influence responses, in particular the extent to which automatic and controlled responses correlate. Notably, however, even seemingly automatic responses are sometimes controllable.

People do differ in their orientation to being egalitarian, and chronically active egalitarian goals sometimes can prevent preconscious, automatic activation of stereotypes (Moskowitz, Gollwitzer, Wasel, & Schaal, 1999); this occurs even when the goal is activated preconsciously, as well as when the stereotype control operates below consciousness (Moskowitz, Salomon, & Taylor, 2000). A parallel set of findings emerges for people who typically feel uncertain about why things happen as they do; chronic causal uncertainty can lead people to have an accuracy goal in understanding other people, and they consequently use stereotypes less (Weary, Jacobson, Edwards, & Tobin, 2001).

A final type of individual control entails people's assessment that they may not be entitled to judge the other person. Social judgeability theory supposes that people do not always judge other people; instead, they must feel entitled to judge (Leyens, Yzerbyt, & Schadron, 1992), based on available data, appropriate roles, relevant values, and so on.

To summarize, motivation and capacity operate in these studies of control over automatic affect-laden attitudes, most of which examine individual differences in motivation through should-would discrepancies, prejudice scales, known groups, or motivations to control prejudice. These studies mostly operationalize capacity by creating stimulus exposures at automatic or controlled levels. Across the board, both capacity and motivation interact with individual differences in motivation to control attitudes.

SITUATIONAL GOALS CONTROL AUTOMATIC ATTITUDES

Just as individual differences allow capacity and motivation to combine in controlling biased responses, so too can situational goals combine capacity and motivation. For example, one set of studies (Blair & Banaji, 1996) compared motivation (intentions: to respond in stereotypic or counterstereotypic ways) and capacity (time: rapid exposures requiring automatic responses versus slower exposures allowing controlled responses). People saw a series of common first names on a computer screen, judging gender as quickly and accurately as possible, by pressing an M or F key on the keyboard. Some received instructions that the name would always follow a stereotypic word (e.g., for males, decisive, crude, jeep, wrestling; and for females, caring, dependent, perfume, laundry), and others learned that the name would always follow a counterstereotypic word. In either case, they could respond more effectively by using the first word as a strategy for responding to the names.

When people intended to respond in a nonstereotypic way and had few capacity constraints, they were able to control their nonautomatic stereotypic associations. Start with  Figure 6.4(A) . In a baseline condition of automatic responding without any strategic intention, people responded faster to male-male matches and female-female matches, as is typical in stereotyping research. They also responded faster to stereotype matches in either automatic [left side of  Figure 6.4(B) ] or controlled [left side of  Figure 6.4(C) ] conditions when using a stereotype strategy.

Figure 6.4  Automatic and Controlled Attitude Processes, as a Function of Intent (Strategy)

Note: Shorter bars indicate faster responses, for example when prime and target gender match. Source: From Blair & Banaji, 1996. Copyright © American Psychological Association. Adapted with permission.

However, when they intended to respond counterstereotypically, they were completely able to do so under controlled conditions [right side of  Figure 6.4(B) ] and were even able to moderate their stereotypic responses under automatic conditions [right side of  Figure 6.4(C) ]. This last point is key: Their intention to respond nonstereotypically allowed them to moderate their associations even when cognitive constraints were high, operating automatically. What's more, in other research, people's motivation and capacity combined to enable them to block not only automatic stereotypic associations but also stereotypic memories (Macrae, Bodenhausen, Milne, & Ford, 1997).

Situational goals can also motivate people to learn new ways of responding. Just as attitudes can form through conditioning, so too can they change through implicit evaluative conditioning (Olson & Fazio, 2006). People can even train themselves, through practice, to “just say no to stereotyping” (paraphrasing an old antidrug campaign slogan). In one study (Kawakami, Dovidio, Moll, Hermsen, & Russin, 2000), people's automatic associations were stereotypic and remained so when they were trained to say no to an irrelevant category. But if they practiced negating a particular stereotypic category, they could control the otherwise automatic activation of skinhead and racial stereotypes. Further, this kind of training against stereotypic associations can carry over to less discriminatory hiring decisions, at least sometimes (Kawakami, Dovidio, & van Kamp, 2005).

Similarly, the goal of perspective-taking can decrease even automatic stereotyping and increase perceived overlap between self and outgroup, as well as evaluations of the outgroup (Galinsky & Moskowitz, 2000). Empathy apparently mediates the relationship between perspective-taking and reduced prejudice (Dovidio et al., 2004; Vescio, Sechrist, & Paolucci, 2003).

Alerting people to their own moral hypocrisy (their general egalitarian values versus specific failures around prejudice) makes implicitly prejudiced people feel uncomfortable and guilty, as well as behave better (Son Hing, Li, & Zanna, 2002). Guilt motivates people to control their prejudice (Amodio, Devine, & Harmon-Jones, 2007). As noted earlier, self-focus also reminds people of their better selves, and they override activated stereotypes, instead of behaving automatically in tune with them (Dijksterhuis & Knippenberg, 2000). And people instructed that “your job is to be as unprejudiced as possible” respond less stereotypically even on automatic reactions (Lowery, Hardin, & Sinclair, 2001).

We have known for a while that goals control less automatic forms of prejudice, as when people are instructed to be accurate (Biesanz, Neuberg, Judice, & Smith, 1999; Biesanz, Neuberg, Smith, Asher, & Judice, 2001; Neuberg, 1989; Neuberg & Fiske, 1987) or color-blind (Wolsko, Park, Judd, & Wittenbrink, 2000). The important insight here is that situational goals, especially given adequate capacity, can reduce even automatic forms of prejudice, an attitude that especially provokes control attempts.

People sometimes attempt to suppress stereotypes altogether, to keep them out of mind. Naturally, stereotype suppression fails when capacity is limited (e.g., by time pressure: Wegner et al., 1992; or by alcohol: Bartholow, Dickter, & Sestir, 2006). And the other necessary ingredient, motivation, appears in several guises—instructed suppression (Macrae, Bodenhausen, & Milne, 1998; Wyer, Sherman, & Stroessner, 1998) and self-focus (Macrae, Bodenhausen, Milne, & Jetten, 1994).

People can be effective when their motivated control is internal, rather than external; they monitor their conflicting responses, as indexed by neural signals for control (Amodio et al., 2008). The mechanisms differ for internal versus external cues to control (Amodio, Kubota, Harmon-Jones, & Devine, 2006). Specifically, although regulating racial responses activates the anterior cingulate cortex, internal cues implicate a more dorsal (upper) region, whereas external cues implicate a more rostral (forward) region. Nevertheless, activating this region does not guarantee success in regulating racial bias, only the intent to try (Amodio, Harmon-Jones, Devine, Curtin, Hartley, & Covert, 2004).

Worse yet, these efforts can backfire. People who actively try to suppress a stereotype may experience afterward a rebound: increased stereotypic responding, relative to baseline participants who never attempted suppression (Macrae et al., 1994; Wegner, 1994). Again, both motivation (goal) and capacity (cognitive load) determine whether rebound follows stereotype suppression (Monteith, Sherman, & Devine, 1998; Wyer, Sherman, & Stroessner, 2000). Lacking motivation, highly prejudiced people may be most vulnerable to rebounds after suppression (Monteith, Spicer, & Tooman, 1998). Indeed, prejudice and goals together determine the degree of rebound after suppression (Monteith, Sherman, & Devine, 1998).

Just as the best intentions can backfire, so too are people's intentions not always good. Other, less benign situational goals also determine biases. Leaving aside overtly prejudiced people, stereotyping and prejudice can result from motivations to support one's desired point of view (Kunda & Sinclair, 1999). For example, students negatively evaluated by a female professor rate her as less competent than a male professor who negatively evaluates them and as less competent than either a male or female who evaluates them positively (Sinclair & Kunda, 2000). Praise or blame by a black doctor has the same effect on white students, especially those high in prejudice (Sinclair & Kunda, 1999).

More generally, a situationally driven threat to one's self-image encourages stereotyping even of those not responsible for the threat (Fein & Spencer, 1997), and this is especially true when people are overloaded (Spencer, Fein, Wolfe, Fong, & Dunn, 1998). Downward comparison and stereotyping (which derogates the outgroup) both can raise self-esteem (Fein & Spencer, 1997). As noted earlier, people with high self-esteem may be more likely to use downward comparisons and outgroup derogation to restore or maintain their self-esteem under threat (Brockner & Chen, 1996; Crocker, Thompson, McGraw, & Ingerman, 1987), a point to which we will return.

Being in a bad mood for whatever reason also can accentuate prejudices and discrimination, especially when the group is personally relevant (Forgas & Fiedler, 1996). When the context puts them in a bad mood, people change the meaning of the outgroup's stereotypic characteristics (e.g., assertivebecomes aggressive; Esses & Zanna, 1995). However, the type of bad mood matters. Being in a sad mood can make people more careful to avoid stereotyping (Bodenhausen, Gabriel, & Lineberger, 2000; Lambert, Khan, Lickel, & Fricke, 1997; Park & Banaji, 2000), whereas an angry mood makes people more likely to stereotype (Bodenhausen, Sheppard, & Kramer, 1994). Conversely, being cheerful can make people careless about using stereotypes (Bodenhausen, Kramer, & Suesser, 1994; Park & Banaji, 2000).

Most often, given current norms, situation-driven motivation takes the form of attempting to avoid biases, and, if so, sufficient cognitive capacity is required. Sometimes, motivations may encourage bias, and then capacity appears irrelevant because the automatic default response and the motivated response both facilitate prejudiced attitudes.

But capacity matters at the earliest processing stages. All the manipulations of situational goals and motivations take as a default the automatic activation of attitudes as soon as people encounter another person. However, the activation and application of stereotypes does differ, depending on capacity (Gilbert & Hixon, 1991). Capacity constraints can undermine the initial activation (accessibility) of stereotypes but can facilitate their application (use) once activated. Given this distinction, goals can interact with either the activation or application stage. The studies described so far usually focus on suppression and inhibition of attitudes otherwise already activated.

However, some goals may intervene at very early stages of activation. For example, people observing faces normally categorize them quickly on the basis of age, ethnicity, and gender. But when people view the faces in a presemantic fashion—for example, search for a scratch on a photographic negative—they may not even see the face as a face. In one pair of studies (Macrae, Bodenhausen, Milne, Thorn, & Castelli, 1997), some people viewed a photograph while searching for a white dot on the surface (presemantic), others while judging whether it was an animate or inanimate object (requiring categorization), and others simply reporting that a stimulus had appeared (presemantic). Only the people who had processed the faces under a categorization task subsequently showed stereotype activation, and this occurred at both longer and shorter processing times, so it was not a function of sheer capacity.

Overall, a variety of situational goals can affect even automatic attitudes (prejudices, stereotypes, discriminatory tendencies), depending on people's available capacity. The goals provide the motivation, and capacity determines how effective it can be. The surprise is that even fairly low-capacity, relatively automatic processes are open to influence.

STIMULUS CONTEXT AS CONTROL

Stimulus context (i.e., the setting in which the target person appears) also determines automatic activation of attitudes. Here, motivation and capacity are less apparent factors than are initial categorization processes. Negative stereotypes and prejudices can dissipate when whites view blacks in a positive context (family barbecue, church) versus a negative one (gang incident, street corner), and these effects operate at automatic levels (Wittenbrink, Judd, & Park, 2001b). Context created by positive and negative exemplars has the same effect on subsequent implicit prejudices (Dasgupta & Greenwald, 2001). An immediately prior context that repeatedly pairs an outgroup with positive associations (Denzel Washington) and an ingroup with negative ones (Jeffrey Dahmer) can attenuate whites' implicit prejudices (Karpinski & Hilton, 2001). Focused mental imagery that is counterstereotypic (e.g., a strong woman) can counteract typical implicit stereotypic associations. So too a black experimenter reduces implicit and automatic prejudices (Lowery et al., 2001). Conversely, repeated associations of a particular person and stereotypic behavior speed up trait judgments, especially for high-prejudice perceivers (Stewart, Doan, Gingrich, & Smith, 1998). Finally, a short-term context that creates arbitrary new ingroups and outgroups can elicit automatic biases (Ashburn-Nardo, Voils, & Monteith, 2001). All these immediate effects of stimulus context apparently operate at relatively early stages of categorization, facilitating some associations over others.

SUMMARY OF CONTROL OVER AFFECT-LADEN ATTITUDES

Capacity and motivation combine to allow individual differences and situational goals to create varieties of control over prejudices. Individual differences in prejudice, motivation to control prejudice, and should-would discrepancies, as well as known group differences, all operate under combinations of capacity and motivation to exert sometimes surprising degrees of control over spontaneous intergroup attitudes. Situational goals to respond counterstereotypically have similar effects. Finally, immediate stimulus context determines degrees of control over relatively automatic attitudes.

Summary of Attitude Formation via Affect

Attitudes can form or change, either affectively or cognitively, and matching seems effective. Nevertheless, the bulk of social psychology's work on attitude formation has had an affective flavor, as we have seen, and the bulk of the work on attitude change has had a cognitive flavor, as we shall see.

To review, attitudes can form through classical conditioning (repetition and association), instrumental conditioning (rewards and punishments), and their social learning counterparts: modeling (imitation) and vicarious reinforcement (observational learning). All these processes can operate with minimal cognitive activity of the opinion-formation variety; instead, they rely on affective processes, such as pleasure and pain. For example, many of people's political opinions are probably based on repetition and association (the smiling candidate appearing in pleasant or important contexts), reinforcement (friends' agreement), modeling (following opinion leaders), and vicarious reinforcement (watching other people get praised or trashed for their attitudes).

Another minimally cognitive and maximally affective process, mere exposure, increases liking via the sheer frequency of encounter with an initially neutral attitude object (the candidate's name on billboards or bumper stickers). The process can operate below awareness. Primary emotional appraisals also occur quickly, preceding more complex emotional reactions. Affective processes appear to be primary and automatic in many respects. Nevertheless, as prejudice research shows, individual differences, situational goals, and stimulus context all can moderate automatic attitudes. Further, attitudes can form either affectively or cognitively, and attitude change may work best when it matches the original basis of the attitude.

Neither conditioned learning nor mere exposure as a basis of opinion formation provides any cognitive content, only affective leanings, which do not exactly build the foundation for an informed democratic citizenry. Nonetheless, even simple, affect-based attitudes serve the object appraisal function of approach and avoidance, contributing to the core motive of understanding and perhaps even democratic expression.

ATTITUDE CHANGE VIA DISCOMFORT WITH CONTRADICTION: UNDERSTANDING AS COGNITIVE CONSISTENCY

Having formed an attitude, however preliminary and merely affect-based, people over time develop cognitions that support that affect. One of the mechanisms for moving from affect-based attitudes to more multidimensional attitudes is probably cognitive consistency. People have a penchant for internally consistent attitudes, as several theories proposed at the end of the 1950s, a decade of convention and consistency. The penchant for consistency fuels the core social motive of coherent understanding and the corresponding attitude function of unambiguous object appraisal.

Theories of Cognitive Consistency

Consistency theories dominated social psychology in the 1960s. Collected in Theories of Cognitive Consistency: A Sourcebook (Abelson et al., 1968)—informally known by its unfortunate acronym TOCCAS—they home in on the cognitive correlates of attitudes. The consistency theories make the simple but profound point that inconsistency begets attitude change. When important cognitions collide, the attitude destabilizes, and some aspect usually caves. Inconsistency can occur within or among the cognitive, affective, and behavioral correlates of attitudes, but most of the theories emphasize cognition and cognitions about behavior.

The experienced inconsistency is psychological, not objective. Thus, if a dieter has salad and a diet drink for dinner, followed by a hot fudge sundae with whipped cream, that objective inconsistency may or may not constitute a felt inconsistency, depending of course on the adroitness of the dieter's ability to rationalize. The comments of outside observers, such as a partner or parent, are psychologically irrelevant, unless adopted as the dieter's own standard.

If the inconsistency is indeed experienced, consistency theories posit the inconsistency to be an unpleasant state of affairs, in need of quick remedy. Just as hunger, thirst, or sexual deprivation causes a drive to satisfy the need, restoring comfort, so too does perceived inconsistency. According to this kind of formula, called a drive reduction model, inconsistency (or hunger, etc.) leads to tension and arousal, which motivates a drive to reduce that discomfort, which encourages remedial activity, in this case, to restore consistency. Examples of consistency models include balance theory (Heider, 1958), which is postponed until the next chapter, and cognitive dissonance theory, addressed next.

Dissonance Theory

People do not believe in cognitive dissonance as an effective way to change attitudes (Snell et al., 1995), perhaps because it is counterintuitive. The best-known theory in social psychology, cognitive dissonance theory has worked its way into everyday language, with all kinds of people claiming to experience “cognitive dissonance” when one attitude is incongruent with another or when they behave incongruently with their attitudes. And, to that degree, people do understand the theory. However, they do not believe that the dissonance they experience could change their attitudes. The details, moreover, explain exactly how the psychology of dissonance does shift attitudes.

THEORY

The inventor of cognitive dissonance theory, Leon Festinger (1957), proposed that cognitions can be relevant or irrelevant, and—if relevant—consonant or dissonant. Dissonance describes both the perceived incongruity and the discomfort predicted to result; people feel tense, aroused, and uneasy from salient, self-involving incongruities. This state of dissonance comes from the ratio of dissonant to consonant cognitions. A typical example includes the cognition “I smoke,” along with all the supporting and countervailing reasons (see  Table 6.3 ; note that smokers naturally muster more consonant than dissonant cognitions).

TABLE 6.3 Cognitive Dissonance in Action: Tyical Reasons for Smoking

Pro	Con
Novel	Nasty
Smoking with friends	Expensive
Relaxing	Deadly
Social	Addictive
Ease of quitting
Weight control

Another, less typical example would be one's active support for a respected, famous person (the President, a Supreme Court justice, a well-known professor), coupled with the knowledge that the person regularly participates in extramarital affairs or sexual harassment. The incongruity causes psychological discomfort and aversive arousal that beg for resolution. People resolve dissonance in several ways, most easily by self-generated attitude change, that is, changing the dissonant cognitions (I can quit smoking any time; the famous sexual harasser's accomplishments outweigh his poor interpersonal judgment).

People change their attitudes and reduce their dissonance specifically by adding consonant or subtracting dissonant cognitions, as well as minimizing the importance of the dissonant ones. Consider an example of subtracting dissonant cognitions (Gibbons, Eggleston, & Benthin, 1997): Smokers who participated in a cessation program fully appreciated the health risks associated with smoking, that is, they accumulated cognitions dissonant with smoking. But if they relapsed afterward, their perceptions of the risks declined, presumably because they subtracted dissonant cognitions about the dangers of smoking. They could not change their behavior, so they changed their cognitions about smoking instead, thereby maintaining consonance and a view of themselves as sensible.

CLASSIC STUDIES

One of social psychology's classic experiments demonstrated attitude change as a function of dissonance between perceived actions and attitudes. The hypothesis was that incongruity between one's attitudes and actual behavior would produce dissonance, which in turn would cause attitudes to change in line with past behavior, which cannot be changed. In the study (Festinger & Carlsmith, 1959), which established the induced compliance paradigm, Stanford undergraduates spent half an hour using one hand to put 12 wooden spools into a tray, empty it, and then refill it again.

The next half-hour's task required using one hand to turn 48 square wooden pegs a quarter turn each, one after another, row by row, and then starting over when reaching the end. After the experiment was apparently over, participants were offered $1 or $20 to tell the next participant that the task was interesting, supposedly to help the experimenter. A control condition received no payment and told no lie.

After telling the lie, the people paid only $1 had considerable dissonance to resolve (the $20 people had a ready explanation for their lie, so no dissonance). The people paid less were more likely to change their own attitudes and rate the experiment as more enjoyable and important ( Table 6.4 ). This readjustment of their attitude resolved their dissonance (not so much pay, but not really a lie). This counterintuitive finding—less incentive causes more attitude change—flies in the face of learning theory predictions, which would predict that bigger incentives cause bigger change. As noted earlier, laypeople believe in learning theory more than in dissonance as a source of attitude change. What's more, people overestimate the power of self-interest (Miller & Ratner, 1998), so the idea that $1 changes an attitude more than $20 remains especially intriguing.

TABLE 6.4 Effects of Cognitive Dissonance on Interview Questions

	Experimental Condition
Question	Control	$1.00	$20.00
How enjoyable tasks were (scale of −5 to +5)	−0.45	+1.35	−0.05
Scientific importance (scale of 0 to 10)	5.60	6.45	5.18
Would participate in similar experiment (scale of −5 to +5)	−0.62	+1.20	−0.25

Source: From Festinger & Carlsmith, 1959. Copyright © American Psychological Association. Adapted with permission.

In the Festinger-Carlsmith experiment, participants' option to resolve dissonance by changing their behavior was blocked. Past behavior cannot be changed. Even continuing behavior resists change more than cognitions do, so attitude change is the easier route. In the case of smoking, one can attempt to quit, thereby resolving dissonance, but changing one's attitudes toward smoking is easier than overcoming the addiction. Consistent with the idea that cognitions are easier to change than behavior, people sometimes misremember their previous behavior as fitting their current attitudes and behavior, thereby changing their cognitions about their prior behavior (Ross, 1989).

Another classic study established the free-choice paradigm (Brehm, 1956). Dissonance theory predicts that when people make a choice, the positive features of the rejected alternatives and the negative features of the chosen alternative provoke dissonance. Hence, people are predicted to bolster their choices, once made, by downplaying these dissonant aspects of the chosen and rejected alternatives. In the context of a marketing study, women rated eight housewares (e.g., toasters, coffeemakers) and could choose to keep one of two then offered to them. The items in the choice pair were selected to be close in desirability (hard choice) or distant in desirability (easy choice). After choosing one, the women again rated the products. Under the hard choice condition, dissonance should be higher, and the women indeed bolstered their choice by spreading the alternatives, that is, liking the chosen one more than before and disliking the rejected one more than before. For the easy choice, with less dissonance, the spreading of alternatives was less, as predicted.

Another classic method of dissonance research, the effort justification paradigm, rests on the prediction that the more unpleasant the attempt to attain a desired goal, the more people should like the goal once it is obtained (Aronson & Mills, 1959). The unpleasant effort is dissonant with the positive features of the goal, and resolving the dissonance requires bolstering the desirability of the obtained outcome. In the classic study, women underwent a screening test to join a group that would discuss sexual topics. That initiation involved reading sex-related words to a male experimenter (supposedly to see whether the women would be able to participate in the group), and the words were either explicitly embarrassing or mildly suggestive. Then the participants listened to a tape of the group having a technical and boring discussion about animal sex. In the severe, high-effort condition, the women had more unpleasant effort to justify, so they reported liking the group more than in the mild, low-effort condition.

One of the obvious predictions from dissonance theory is selective perception; that is, people should be prone to heeding, interpreting, and remembering information that supports their beliefs, thereby adding to their consonant cognitive reserves. Although a principle that deserved to be true, early research found that people do not actively avoid dissonant information, but in the course of their lives, they tend de facto (by default) to be exposed to more information that supports their views (Freedman & Sears, 1965). Further, notions of utility, fairness, and novelty can override the tendency to avoid views that disagree with one's own. Under some conditions, some versions of dissonance do affect early perceptual processes. Participants who thought they freely chose to wear a flamboyant tropical costume, while crossing campus, or to push themselves uphill, while kneeling on a skateboard, reported that the environment was less aversive than did low-choice participants performing the same embarrassing feats (Balcetis & Dunning, 2007); dissonance effects on early perception explain these results.

The entirely compatible tendency toward seeking supportive (consonant) information (rather than actively avoiding dissonant information) was pointed out later (Eagly & Chaiken, 1998; Frey, 1986). Selective exposure to supportive views occurs when people have freely chosen the dissonant activity and are publicly committed, but also when people's attitudes are moderately stable, neither so tentative that they are seeking the widest range of information, nor so fixed that they are checking for dissonant information that might be useful in the long run. High authoritarians (who have rigid attitudes), when under threat (from thinking about their own death), especially show selective exposure (Lavine, Lodge, & Freitas, 2005), suggesting its utility for people with defensive, protective stances and low tolerance for the ambiguity of opposing views.

Similarly, selective interpretation allows people to evaluate unfavorably any information that might dispute their views. Recall from  Chapter 1  the doomsday cult that reinterpreted information to fit their shared understanding, especially when doomsday failed to arrive as prophesied (Festinger, Reicken, & Schachter, 1956).

However, evidence of selective memory for supportive information is less clear (Eagly, Chen, Chaiken, & Shaw-Barnes, 1999); for example, strong opinions on contentious subjects allow people to remember their own side, but also the other side. Moreover, amnesiacs as well as participants with normal memory but distracted by cognitive load (Lieberman, Ochsner, Gilbert, & Schacter, 2001) still show dissonance reduction, suggesting that it does not require explicit memory at all. Thus, the evidence for selective exposure and interpretation is stronger and clearer than the evidence for selective perception and memory. As in social cognition, most of the interesting processes occur online (in exposure and interpretation), rather than on the basis of memory.

CONTROVERSIES

Dissonance theory provoked some controversy in its heyday (Aronson, 1992; Harmon-Jones & Mills, 1999; Petty & Wegener, 1998). The first challenge, self-perception theory (Bem, 1967; recall from  Chapter 3 ), held that people infer their attitudes from observing their own behavior and its context. Therefore, when people observe their own effort, choices, or incentives, they infer the corresponding attitudes. No sense of dissonant discomfort or changed cognitions is necessary, in this view. Another challenge also rejected the idea of dissonance-provoked tension, arguing that the findings all result from impression management, that is, participants trying to appear rational and consistent to the experimenter (Tedeschi, Schlenker, & Bonoma, 1971), suggesting a role for self-enhancing or belonging motives.

In response, the ongoing debate centered around the hypothesized state of dissonance, the discomfort itself. If people experience psychological discomfort or physiological arousal from dissonance, then the self-perception and impression management explanations are less plausible. Indeed, dissonance does elicit tension, measurable discomfort, and unpleasant physiological arousal, thereby supporting the original theory as an explanation (for reviews, see Cooper, 1995; Harmon-Jones & Mills, 1999). Mere impression management seems even less likely, considering that dissonance activates left frontal brain regions that correlate with questionnaire measures of dissonance-based attitude change (Harmon-Jones, Harmon-Jones, Fearn, Sigelman, & Johnson, 2008).

Recent variants on the theory abound (Harmon-Jones & Mills, 1999). According to the self-standards model (Stone & Cooper, 2001), all variants concern people violating self-standards, either personal or normative. Personal standards suggest a role for self-enhancing motives, and normative standards for belonging motives. Combined, as in public advocacy and private reminders of one's failures to live up to self-standards, dissonance can change behavior toward healthy, environmentally friendly, and interpersonally responsible behavior (Stone & Fernandez, 2008). In other more specific models, the self fails either private or public standards, as follows.

If people's behavior violates norms, that is, external self-standards, then self-esteem plays no role in dissonance reduction. Under these circumstances, described by the new look at dissonance theory (Cooper & Fazio, 1984), dissonance may simply require people to believe that they have freely chosen an action with foreseeable aversive consequences. To experience dissonance, people essentially have to feel guilty. Felt responsibility for aversive consequences can provoke dissonance, even when one's behavior is consistent with one's beliefs, because the aversive consequences are inconsistent with one's socially established preferences (Scher & Cooper, 1989). The relevant social norms (self-standards) are external to the self, so self-esteem is irrelevant.

Alternatively, any action may produce dissonance if it violates one's self-consistency, that is, personal expectancies about one's own competent and moral behavior. The higher one's self-esteem, the higher one's personal expectations, and the more vulnerable to dissonance one would be. In this view, people change their attitudes to maintain self-esteem (Aronson, 1968, 1992). This emphasizes one's personal self-standards.

Finally, people may violate personal standards in the context of a broader view of self. If one has abundant, accessible, positive cognitions about self, and if one has violated a personal standard, the repair work is easy. Thus, people with high self-esteem would be less vulnerable to dissonance. In this broader view, dissonance results from threats to self-integrity, leading people to affirm their worth in a variety of ways, which include but are not limited to attitude change (Steele, 1988). In this view, the self is a resource enabling self-affirmation in the face of dissonance. Dissonance persists when people attempt self-affirmation but fail (Galinsky, Stone, & Cooper, 2000).

Summary of Attitude Change via Discomfort

Consistency theories focus on people's tendency to maintain clusters of coherent attitudes, cognition, affect, and behavior. Operationally, tests of the theories have emphasized people's cognitions about their actions and their beliefs. Cognitive dissonance theory shows that people seek to keep their cognitions consistent, including their cognitions about their attitude-relevant behavior. Thus, cognitive consistency theories partly illustrate the core social motives of self-enhancing and belonging, seeing oneself as not violating personal or normative self-standards. Nevertheless, the core social motive of understanding primarily underlies consistency processes, in that people seek the clear object appraisals afforded by cognitively consistent attitudes. Most dissonance research emphasizes self-inflicted attitude change; we now turn to attitude change engineered by others.

ATTITUDE CHANGE VIA UNDERSTANDING PERSUASIVE COMMUNICATION

Recall the corporate merger that opened this chapter. The public relations persuasion campaign consumed millions of dollars and person-hours. Persuasion theories focus on just such deliberate attempts to change people's attitudes, as in advertising, propaganda, legal argument, and health education. People think they can resist persuasion (Wilson, Houston, & Meyers, 1998): People say they are unwilling to change their attitudes, think they can generate effective counterarguments, and later underestimate how much their attitudes indeed have changed. Social psychologists know better. Researchers analyze persuasive communications for their effectiveness, typically by breaking down the process, as this section will show in two major approaches, one classic and one new. The object-appraisal function, exemplifying the core social motive of understanding, especially represents how persuasive communications succeed. Note the general focus on cognitive processes of persuasion, as compared with the more emphatically affective processes of attitude formation described earlier. The persuasive communication theories take more account of people's abilities to understand issues and ideology, which indeed predict important behavior, such as voting (Jost, Federico, & Napier, 2009).

Yale Communication and Persuasion Approach

Persuasive communication analyses proceed by dissection. Breaking processes down into their components, these theories achieve precise and rigorous enumeration of all the likely relevant variables, just as any dissection lays all the functional parts on the table.

MCGUIRE'S INFORMATION-PROCESSING STAGES

Consider a theory that slices persuasion according to audiences' information-processing stages. William McGuire (1969) emphasized people's initial reception of messages, as comprising exposure, attention, and comprehension. For example, in a political campaign, if aspirants do not get the message out, they cannot be effective. Exposure, as in distributing campaign messages via 30-second television spots, famously requires resources (money, staff). But no amount of exposure is effective if people do not attend, instead using the television spot to take a nacho break. And comprehension, of course, must follow attention: If the ad is confusing or obscure, the audience cannot take it in. Message reception matters most when it is problematic (Eagly & Chaiken, 1993): when the audience is strongly distracted, verbally unintelligent, or receiving a difficult message in an audio or video modality (as apart from written messages, which can be read at one's own pace). Early researchers measured reception only as memory for the message, but subsequent researchers have realized that a recipient could form an attitude online, retaining the attitude but forgetting the details of the message. In that case, memory for the message would not appropriately measure reception (Chaiken, Wood, & Eagly, 1996).

Once the message is received, successful persuasion requires yielding, that is, accepting and using the message. The major stages—reception and yielding—can be sensitive to different processes, so they are useful to separate. For example, self-esteem effects on persuasion are independently explained by both message reception and yielding (Rhodes & Wood's meta-analysis, 1992): People low in self-esteem have difficulty receiving a message, perhaps because they are too distracted and withdrawn; they indeed fail to recall messages. At the other end, people high in self-esteem do not yield easily because they are especially confident in their opinions. The net effect is that people of moderate self-esteem are most easily persuaded because the reception and yielding processes both cooperate.

McGuire broke the yielding process down into its components: yielding (the actual agreement), retention (keeping and remembering the changed opinion), and acting on it. Other theories of persuasion and attitude-behavior relations more fully address these latter stages, so we now turn to those.

COMMUNICATION AND PERSUASION VARIABLES

World War II brought many psychologists into the military, and among them were those studying enemy propaganda. These studies resulted in a postwar Yale research laboratory including Carl Hovland, Irving Janis, and Harold Kelley. Theoretically eclectic, their research systematically examined each component of persuasion, breaking the interpersonal process down according to a well-known formulation by Harold Lasswell (1948): Who says what to whom with what effect. Their research program focused on these variables, more than on theoretical analysis: source, message, recipient, and context. For example, features of the source influence the success of persuasion, for whether the source possesses credibility (e.g., expertise and trust) affects persuasion.

The feature of the message most often studied is the strength of the argument. As one classic example, the Yale laboratory's research program examined the impact of fear appeals (Janis & Feshbach, 1953), that is, messages that scare people. Researchers presented participants with tooth-brushing messages, embedded in varying degrees of threat. The most severe fear appeal contained 71 separate references to negative consequences of dental neglect, from tooth decay to gum disease to oral cancer, coupled with vivid photographs of same, and couched in accusatory language (“This can happen to you”). Low levels of fear were created by minimal references to mild negative consequences of neglect, primarily tooth decay, coupled with photographs of healthy teeth, and couched in impersonal language. The two fear appeal groups and a control group (who received an irrelevant message about the human eye) were asked a week later how much they had complied with various dental hygiene recommendations. Only the low-fear appeal was effective, whereas the high-fear appeal did not differ from the control message.

These results suggested that too much (or too little) fear undermines persuasion, a result again explained by McGuire's emphasis on reception versus yielding processes, just covered. McGuire suggested that (like self-esteem) fear has opposite effects on the two processes: Fear blocks reception but enhances yielding. Adding the two processes together, the lowest levels of fear should allow reception but not motivate yielding, whereas the highest levels would allow yielding but block the initial reception. Moderate fear balances the ill effects on reception with the beneficial effects on persuasion, in this view. Thus, the effect of fear on persuasion should be an inverted U, with most persuasion at moderate levels ( Figure 6.5 ).

Figure 6.5  Effects of Fear Levels on Message Reception, Yielding, and Ultimate Persuasion

Source: Drawn from model described by McGuire, 1969.

The problem with an inverted-U prediction is that researchers must sample the entire range of the predictor variable, in this case fear. If a researcher samples only the low-to-moderate end of the range, the observed relationship will be positive: Moving from low to moderate fear increases persuasion. If a researcher samples only moderate-to-high levels of fear, then the observed relationship will be negative: Moving from moderate to high fear decreases persuasion. The Janis and Feshbach study, which showed a negative relationship, may not have sampled enough of the low-to-moderate end of the range. Most of the fear appeals research does show that threat enhances persuasion (Eagly & Chaiken, 1993; Witte & Allen, 2000), but maybe most of the recent research neglects the graphic, high-level fear appeal invoked by the original study, so it does not show the curvilinear effect. People's perceived vulnerability and outcome severity also interact with fear effects (de Hoog, Stroebe, & de Wit, 2007), so fear is not a simple variable.

We have illustrated communicator and message variables; the next two sections address recipient and context variables in discussing two later models of persuasion processes. Overall, the contributions of the Yale communication and persuasion approach are considerable, although many specific findings have been qualified in subsequent research. Essentially, this single persuasion laboratory set the scientific agenda for subsequent researchers, establishing persuasion as a topic to study, specifying the important variables, and demonstrating experimental rigor. However, the resulting research perhaps overemphasized deliberate, thoughtful persuasion processes, based on careful processing of message content. Other persuasion processes—considerably more superficial—matter as well.

Dual-Process Persuasion Models

People search for understanding and specifically object appraisal in different ways at different times. In the social cognition chapter, more automatic versus more controlled processes divided, respectively, the rapid, expectancy-based, heuristic processes from slower, more detailed deliberate processes. Simultaneously, attitude theorists had identified persuasion operating via two routes, one fast and superficial versus one slow and analytic. The heuristic-systematic model and the elaboration likelihood model both make this point.

HEURISTIC-SYSTEMATIC MODEL

Imagine that your university contemplates switching from its present two-semester system to a trimester system, but not for at least another five years. The change advocate is a cynical administrator who views students as irresponsible and immature, as unconcerned with their societal role, and as getting more public credit than they deserve. The proposed change would mean 50% more courses, exposing students to a wider variety of material, and each course would cover 70% of the standard material. The year would start later, and the first term would still end before winter holidays. Many schools use this trimester system and view it as advantageous to their graduates. To what extent do you think carefully about this issue? If the change is delayed for more than five years, it seems remote and irrelevant. And the fellow pushing it sounds like a jerk, so it is probably a bad idea.

On the other hand, suppose you hear advice that it is actually healthier to sleep less than eight hours a night, and soon you will have to be interviewed about your opinion on this topic. The less-sleep position, advocated by a well-loved campus figure, includes arguments such as the cultural and seasonal nature of sleep patterns, the efficacy of REM (dreaming) sleep over non-REM sleep, the utility of naps and meditation, correlations between excessive sleep and illness, as well as famous high achievers who sleep less than average. To what extent do you think carefully about this issue? In this case, given that it affects you, and that you soon will have to have a considered opinion, the likeability of the person who conveyed this opinion matters less than the quality of the arguments.

Using a refined version of this methodology, Shelly Chaiken (1980) proposed that (a) highly involved people use systematic processes to understand a persuasive message, deciding its merits by evaluating the arguments, whereas (b) uninvolved people use heuristic processes, simple decision rules such as the likeability of the communicator, to give a more superficial opinion. Systematic, message-based persuasion persists longer than heuristic, superficial persuasion. This work generated the heuristic-systematic model, which proposes that persuasion sometimes operates by simple and effortless shortcuts, while at other times it operates with more careful thought.

A subsequent research program (Chen & Chaiken, 1999) identified sample heuristic shortcuts for superficial persuasion:

· Length implies strength (a message with many arguments is probably right).

· Consensus is correct (if a lot of people agree, then it must be true).

· Go along to get along (whatever your conversation partner asserts is acceptable).

· Experts can be trusted (expert sources are accurate).

These heuristics save mental time and energy, triggering persuasion when capacity and motivation are low. People with limited time or knowledge have reduced capacity for thoughtful processing of persuasive messages, so they rely on heuristics. People motivated to be accurate and confident, in contrast, increase their effort toward systematic processing. In the short term, in the rush of events, people may be distracted by superficial heuristics, focusing on secondary characteristics, whereas when thinking into the distant future, people are more able to focus systematically on primary characteristics and abstract principles (Fujita, Eyal, Chaiken, Trope, & Liberman, 2008).

Heuristic and systematic processes do not constitute a dichotomy or even opposite extremes; they can occur simultaneously (Chen, Shechter, & Chaiken, 1996). Researchers primed either an accuracy motive (systematic) or social-impression motive (heuristic) by having participants read scenarios that all emphasized either objectivity (systematic accuracy) or social norms (heuristic impressions). As the first line of  Table 6.5  indicates, impression-minded participants then simply agreed with their partners (a heuristic process; in effect, “go along to get along”), whereas accuracy-minded participants were unaffected by their partners' opinions (thereby not showing any heuristic process). Similarly, impression-minded partners' systematic thoughts about the message also were biased by their partners' views (see the second line of  Table 6.5 ), whereas the thoughts of accuracy-minded participants were not significantly affected by their partners' views. Thus, for the impression-minded participants, both heuristic processing (agree with partner) and systematic processing (use message content) operated simultaneously.

TABLE 6.5 Heuristic and Systematic Processing: Own Favorability under Impression and Accuracy Motives, Depending on Partner Attitudes

	Motivation
	Accuracy		Impression
	Partner attitude		Partner attitude
Measure of own favorability	Favorable	Unfavorable		Favorable	Unfavorable
Attitudes	7.60	7.22		9.56	5.38
Thoughts	−0.41	0.32		1.01	−0.28

Source: From Chen et al., 1996. Copyright © American Psychological Association. Adapted with permission.

In general, the heuristic-systematic model proposes two parallel routes to persuasion, detailing the heuristic route in particular, as well as demonstrating some of the interpersonal motivations that encourage one or the other type of processing. Note that the interpersonal motives partly illustrate the Yale approach's context variable. Note too the similarity of the accuracy motive to the core social motive of understanding and the resemblance of the impression motive to the core social motive of belonging.

ELABORATION LIKELIHOOD MODEL

At about the same time that the heuristic-systematic theory emerged, Richard Petty and John Cacioppo (1981) also developed a model of persuasion by two routes: peripheral and central. The peripheral route, like Chaiken's, includes heuristics but also other potential shortcuts, such as mood. The central route, like Chaiken's, involves careful (though not necessarily accurate) processing of message information. The difference from Chaiken's heuristic-systematic model is that the elaboration likelihood model operates at a broader level, less focused on specific heuristic and motivational processes.

The elaboration likelihood model posits the probabilities of elaborating (considering and extending) message content, hence the name of the model. Stated simply, various persuasion variables (source, message, context, and recipient), which are familiar from the Yale approach, act to increase or decrease the recipient's likelihood of thinking about the persuasive communication. The valence (pro or con) of these thoughts then determines persuasion.

The difference from the Yale approach is that the elaboration likelihood model (like Chaiken's heuristic-systematic model) examines online message processing, whereas the Yale model had focused on learning and recall of message content. That is, the Yale model had assumed that, to the extent people first learn message arguments, they are persuaded. Thus, the research typically tested people's memory for message content, as an index of message learning, to see if it predicted persuasion. In contrast, the more recent models find that people's online responses—their thoughts as they are receiving the message—more reliably predict their persuasion. Literal memory for message content (which is determined by extraneous factors) is not particularly associated with persuasion. The elaboration likelihood model's central point remains that degree and type of thinking mediate (intervene) between the persuasion attempt and potential attitude change ( Figure 6.6 ).

Figure 6.6  Elaboration Likelihood Model

Source: From description by Petty & Cacioppo, 1981.

One test of this model, then, compares people who like to think with those who do not. People differ in their need for cognition, that is, the extent to which they enjoy and tend to engage in cognitive effort; it is measured by a scale including items such as “I would prefer complex to simple problems”; “I find satisfaction in deliberating hard and for long hours” versus “Thinking is not my idea of fun”; “I only think as hard as I have to.” Over 100 studies on need for cognition (Cacioppo, Petty, Feinstein, & Jarvis, 1996) find that individuals high in need for cognition indeed do think more than those who do not. All else being equal, then, their likelihood of elaborating on message content should be higher than people who do not like to think. For example, people high in need for cognition discern argument quality more than those low in need for cognition. In one study (Smith & Petty, 1996), people high in need for cognition markedly differentiated strong and weak arguments for taking a fictitious Vitamin K, and they were more persuaded only if the arguments were cogent.

Besides need for cognition, other recipient variables include gender, age, self-esteem, intelligence, self-monitoring, knowledge, attitude accessibility, and involvement. As an example of involvement effects, a meta-analysis (Johnson & Eagly, 1989) found that recipients highly involved in a message, either because it was relevant to their enduring values or because they cared about the impression their attitude would make on others, were less easily persuaded than uninvolved recipients. Recipients involved in an outcome relevant to the persuasive message (e.g., a policy change affecting one's own academic fate) differentiated more carefully between strong and weak arguments than uninvolved recipients, a finding consistent with the elaboration likelihood model.

The Vitamin K study illustrates not only a recipient variable (individual differences in need for cognition) but also a message variable. Messages framed negatively (“Not taking Vitamin K can lead to illnesses that shorten life span”) versus positively (“Vitamin K helps you live a longer, healthier life”) elicit more thought and persuasion when the frame is unexpected—but only for those low in need for cognition, because the high-need-for-cognition individuals are already thinking as much as they can. Other message variables, besides framing, include the message's relevance to the recipient and the agreeability of its position. Also, its format matters: whether or not it uses rhetorical questions, draws conclusions, presents both sides, scares people, or invokes other emotional appeals. The most important single factor, of course, is whether it argues strongly. As to which persuasion technique is most effective, the simple answer is: It depends. It depends on the recipient, the context, and the source. Persuasive messages have to be tailored to particular settings and purposes. Small wonder Daimler-Chrysler's highly effective, carefully crafted advertising blitz cost $25 million.

To understand the complexities of crafting a persuasive appeal, and the importance of context variables, consider the effects of mood on persuasion. When a situation makes them happy, people do not like to think, if it will interfere with their mood, but they do like to think, if it will reinforce their good mood. So, they may avoid thinking about doom-and-gloom messages, but be happy to think about a heart-warming one. And if they do think, persuasion depends not only on the quality of the arguments but also on the positive associations made available by the positive mood. Thus, mood has multiple roles in persuasion (Petty & Wegener, 1998). Other context variables have prominently included distraction and time delay between message and measure.

Source variables have prominently featured the communicator's credibility and attractiveness. Using communicator attractiveness reflects a more peripheral route to persuasion, whereas credibility could motivate either a peripheral route (this expert must be right) or a central route (it is worthwhile to think about an expert's arguments). These source variables, along with recipient, message, and context variables, reflect the range of topics addressed by the elaboration likelihood model. The model potentially helps explain the processes by which each variable affects persuasion. As noted, each variable (e.g., mood) can have multiple effects by different processes in different situations (Petty & Briñol, 2008), so unpacking the circumstances and processes matters a lot.

Summary of Persuasive Communication

Persuasion research examines deliberate attempts to change another person's attitude. The earliest work, coming out of wartime propaganda research, established persuasion as an area of rigorous scientific study, laid out the relevant variables, and set the research agenda. Although it emphasized effective processing (i.e., learning) of messages, more recent models emphasize alternative routes to persuasion, based on relatively superficial or in-depth processing. While the broad generalizations are few in this area of knowledge, because persuasion variables have multiple effects, the fine-grained knowledge of persuasion techniques is well developed. Research in this area tends to emphasize the core social motive of understanding when it describes systematic information processing. Other core social motives, such as belonging, appear when research turns to heuristic processes, such as going along to get along. In either case, attitudes matter most when they guide behavior, which occupies the next section.

WHEN AND WHY ATTITUDES MATTER: PREDICTING BEHAVIOR VIA UNDERSTANDING AND BELONGING

If a lawyer persuades a judge, an advertiser persuades a consumer, or a politician persuades a voter, but the judge does not judge, the consumer does not consume, or the voter does not vote, what use is the attitude change without the behavior? Social psychologists at first assumed that attitudes do predict behavior; it seemed so obvious. But a classic study and a later literature review raised the tough question that maybe attitudes do not always predict behavior after all. If attitudes do not predict behavior, then maybe they are mere epiphenomena—by-products of more substantial processes, mere passing trifles of no intrinsic interest. If so, why would researchers research them, pollsters poll them, and the rest of us express them?

Attitudes Don't Always Predict Behavior

In 1930, social psychologist Richard LaPiere approached a hotel clerk, “with some trepidation,” to obtain rooms for himself and a young Chinese couple. To his great relief, given the small town's reputation for bigoted attitudes toward Asians, they were accommodated without a murmur. Passing through two months later, he phoned the same hotel, out of curiosity, asking if they would lodge “an important Chinese gentleman”; they declined. Struck by this discrepancy, LaPiere began keeping track of their treatment as he and the couple together embarked on a 10,000-mile cross-country car trip. In what was to become a citation classic, LaPiere (1934) documented their treatment. As they stopped for 184 meals and 66 overnights at hotels, auto camps, and tourist homes, they were received graciously in all but one instance, when the owner of a dilapidated car camp refused to lodge them, referring to the Chinese couple as “Japs.” Six months later, LaPiere sent a questionnaire to the same establishments, asking whether they would accept “members of the Chinese race” as guests. Fully 92% claimed they would not, again a marked discrepancy from their face-to-face behavior. The study is methodologically flawed; for example, the same people might not have made the face-to-face decisions as answered the letters. Nevertheless, for years, the study was used as a caveat, warning about limitations to the predictive power of survey research, but it hardly dampened researchers' enthusiasm for measuring and reporting attitudes.

Then, in the late 1960s, a review of 47 empirical studies of attitudes and behavior (Wicker, 1969) also concluded that attitudes barely predict behavior, usually correlating less than .30. The studies included job attitudes and absenteeism among aircraft plant employees, young mothers' attitudes and behavior about breast-feeding, plus students' and businessmen's attitudes and behavior toward football, movies, career, and sleep. The correlations were quite variable, but frequently low. Consequently, Wicker dismissed the significance of attitudes: “Taken as a whole, these studies suggest that it is considerably more likely that attitudes will be unrelated or only slightly related to overt behaviors than that attitudes will be closely related to actions” (p. 65). What ails attitude-behavior relations? Several diagnoses emerged, concerning theory, variables, and measurement.

THEORY

First, the theory can be at fault. Researchers had assumed that attitudes cause behavior, but dissonance theory teaches us that behavior can cause attitudes. (Recall that counterattitudinal behavior can shift attitudes.) And self-perception theory (Bem's theory, noted in  Chapter 3 ) teaches us that people's behavior can lead them to infer their attitudes, again making the behavior precede the attitude (at least in conscious form). If the behavior does predate the attitude, the attitude could be a mere epiphenomenon, insignificant in its own right.

OTHER VARIABLES

Another diagnosis of the attitude-behavior problem is that other variables, besides attitudes, influence behavior. As the social influence chapter demonstrates, and as the core motive of belonging reminds us, people's behavior follows norms—the unwritten rules, standards, and values of their social groups—sometimes minimizing the role of personal attitudes. Not only do norms influence behavior, but norms influence the attitudes that people express. Thus, people's so-called attitudes on a survey may not be their only attitudes, and again have little bearing on their behavior. The problem of other variables, beyond the direct attitude-behavior relation, complicates prediction.

MEASUREMENT

Finally, measurement of attitudes and behavior can be faulty. The test of the theory—that attitudes cause behavior—is only as good as the operationalizations of attitude and behavior (as  Chapter 2  argued, this holds for all theories and operationalizations). Two kinds of measurement issues come to the forefront. First, different measurement methods have different biases. For example, self-reports (as opposed to observer ratings) tend to reflect people's efforts to look good. Thus, when self-report methods measure both the behavior and the attitude, attitude-behavior relations are stronger (Kraus, 1995), probably because people present themselves as more consistent than they are or perhaps because their own definition of the behavior corresponds more closely to their own attitude toward the behavior, compared with an outside observer's definition of the behavior.

With regard to a second measurement issue: A general attitude toward, say, the environment, may not predict the specific behavior of recycling scrap paper when printing a late-night essay for psychology class ( Table 6.6 ). The mismatch between general attitudes and specific behaviors suggests either making the attitude more specific (“What is your attitude toward recycling paper from school work?”) or making the behavior more general (a combination of answers to the questions “To what extent do you recycle newspaper, office paper, containers? To what extent do you use public transport, walk, or bike, instead of using a car? To what extent do you conserve water?”). As the next section indicates, matched levels of analysis (specific attitude, specific behavior, or general attitude, general behavior) allow more accurate prediction (Ajzen, 1987).

TABLE 6.6 Match between Attitudes and Behavior

Attitude	Behavior
General: support for environment	Behavioral index, e.g.,
	Recycle newspapers
	Recycle office paper
	Recycle bottles and cans
	Use public transport
	Bike
	Walk
	Conserve water
Specific: support for recycling paper from schoolwork	Recycle paper from schoolwork

In short, the sorry state of attitude-behavior correlations as of 1969—and whenever researchers fail to heed the lessons learned—probably reflects various factors: theory, omitted other variables, and measurement.

When Attitudes Do Predict Behavior

Wicker's pronouncement provoked a response. Researchers set to work to prove him wrong, developing more precise measurements and also more careful theories, which specified that the attitude-behavior relation depends on variables related to the attitude, the situation, and the person.

IT DEPENDS ON PRECISION MEASUREMENT

Icek Ajzen and Martin Fishbein (1973; Fishbein & Ajzen, 1975) proposed the theory of reasoned action to account for careful prediction of behavior from attitudes. Their theory holds that attitudes and subjective norms predict behavioral intentions, which are the best predictors of behavior. Hundreds of studies have assessed the correlations between intentions and actions, finding them to be high on average (.44 to .62). And the prediction of intentions from attitudes toward the behavior and from subjective norms, combined, is also high (.63 to .71), according to a summary of various meta-analyses (Sutton, 1998). Widely applied, the theory predicts, for example, people's use of condoms, dental floss, and doctor visits, among other health behaviors.

The success of this theory depends on accurate measurement. Attitudes are assessed by asking respondents for their evaluations of every aspect of the attitude object: for example, all the perceived consequences of voting for a particular presidential candidate. Once respondents provide all their subjective values (perceived consequences) associated with the attitude object (excellent for the economy; good at international relations; not so good at dealing with Congress), they also provide the subjective likelihood (perceived probability) for each (how certain they are that the candidate will be excellent for the economy). The subjective probability and value are multiplied (to give the more certain values more weight than the uncertain ones) and summed, to create an overall measure of attitude. Subjective norms are assessed in the same way: what various important people are perceived to think of the respondent voting for this candidate and how certain the respondent is about their opinions.

Finally, a third variable, besides attitudes and norms, was added by Ajzen in an updated theory of planned behavior (Ajzen, 1991; Madden, Ellen, & Ajzen, 1992). The new predictor variable was perceived behavioral control, the probability that one can perform the behavior in question (the respondent's estimate of likely control over acting on the intention, once formed). Together, attitudes, subjective norms, and perceived behavioral control predict intention to perform the behavior, which in turn predicts behavior (Ajzen, 2001).

This theory makes explicit the nature of the attitude that best predicts behavior. One could best predict voting for a particular candidate by the person's attitudes toward voting for that particular candidate. A less good predictor would be the general attitude toward the candidate, without any mention of voting. And still less good would be general attitude toward the candidate's political party, absent any specific application to this candidate in this election. The theory also makes explicit the other variables that interfere between attitudes and intentions, namely, what other people are perceived to think and how much personal control one perceives over the behavior. With careful theory, careful specification of the other variables, and careful measurement, especially matching of the specific attitude to the specific behavior, attitudes successfully predict behavior.

One everyday parallel, noted in  Chapter 4 : When people report their specific intentions for implementing behavior, as opposed to merely having a general goal, they are more likely to carry out their intentions (Gollwitzer & Brandstätter, 1997). As noted, difficult personal projects (e.g., write a paper, find an apartment, settle a conflict) were fully three times more likely to be completed if students spontaneously provided implementational intentions, that is, had thought of specific actions in response to certain situational contexts (“when I am in situation y, I will perform behavior z”). Again, specificity matters not only in research but also in life, for attitudes and intentions to predict behavior.

IT DEPENDS ON THE ATTITUDE

Some attitudes predict behavior better than others. In 88 attitude-behavior studies, various variables moderate (qualify) the attitude-behavior relation (Kraus, 1995): Attitudes that are certain, stable, consistent, accessible, and based on direct experience predict behavior the best. In addition (Petty & Wegener, 1998), attitudes high on issue-relevant knowledge and personal relevance also predict behavior. In other words, if people are certain, knowledgeable, and involved in an attitude that is stable over time and place, based on mutually consistent thoughts and feelings, based on direct (rather than secondhand) experience, and mentally accessible at the time of their behavior, all these aspects of the attitude will contribute to its predicting behavior. Essentially, the overriding principles are that highly accessible attitudes predict behavior (Ajzen, 2001; Fazio, 1990) and strong attitudes predict behavior (Ajzen, 2001; Petty & Krosnick, 1995). Strong attitudes are stable and consequential (Krosnick, Boninger, Chuang, Berent, & Carnot, 1993).

As an example of accessible attitudes predicting behavior, researchers (Regan & Fazio, 1977) found that direct experience leads to highly accessible attitudes. A temporary housing crisis at Cornell University demonstrated the importance of direct experience on attitude-behavior correlations. One fall, as happens occasionally at every college, the number of entering first-year students outstripped the number of dormitory beds available. Some students had to sleep on cots in dorm lounges for as long as two months while the situation was being resolved (an example mentioned in  Chapter 1 ). The researchers measured attitudes about the crisis (how tolerable it was; how much suffering it caused; perceived administration efforts, concerns, and priorities) and behavior (the number of actions students agreed to take to resolve the crisis, from signing a petition to joining a committee working on the problem). Students in temporary housing, that is, those with direct experience and presumably accessible attitudes, showed attitude-behavior correlations of about .30, whereas other students—who knew about the problem but had no personal experience with it and held less accessible attitudes—showed attitude-behavior correlations of about zero.

Attitude strength includes many factors (Krosnick et al., 1993). For example, unambivalent attitudes are stronger than ambivalent ones. That is, people's emotions can predict their behavior better than their beliefs do, at least when heart and mind conflict. Voters with unconflicted thoughts and feelings about presidential candidates, that is, unambivalent voters, had internally consistent attitudes that predicted behavior fairly well, and equally from both heart and mind. In contrast, ambivalent, conflicted voters were more likely to vote with their heart than with their mind (Lavine, Thomsen, Zanna, & Borgida, 1998). Internal consistency strengthens an attitude, but absent consistency, affect seems stronger, as we saw earlier.

Attitudes can be strong because of internal consistency but also because of their extremity (Prislin, 1996). This is what laypeople normally consider to be feeling strongly about something: having an extreme viewpoint. More generally, strong attitudes are embedded in one's self-concept, personal importance, values, and knowledge, but they also represent a commitment: extremity, resistance, and certainty (Pomerantz, Chaiken, & Tordesillas, 1995). Strong attitudes resist persuasion and predict behavior.

According to Kraus (1995), in general, correlations with behavior for strong, accessible attitudes average around .50, whereas the same correlations for weak, inaccessible attitudes are less than .20, a substantial difference between a small correlation and a large one, by social science standards. Overall, studies on features of the attitude itself summarize easily: Strong, accessible attitudes predict behavior better than weak, inaccessible ones do.

IT DEPENDS ON THE SITUATION

Some situations emphasize going along with social demands, so they undermine attitude-behavior correlations. In those cases, norms predict behavior better than individual attitudes do. For example, attitude-behavior correlations for reported attitudes toward minority groups show low correlations with actual behavior (.24), presumably because of norms to appear unprejudiced, whatever one's personal views. In contrast, voting and contraceptive use—both behaviors that occur in private—show high attitude-behavior correlations (both .58). Kraus (1995) indicates that attitude-behavior correlations vary a lot, depending on circumstances.

Other situations emphasize the importance of acting on one's attitudes. An experimental demonstration of making one's attitude relevant comes from a classic study (Snyder & Swann, 1976). Male undergraduates rated their attitudes toward affirmative action as part of a larger set of questionnaires, and then two weeks later they came into the laboratory to participate in a mock court case, in which they had to judge liability in a sex-discrimination case. The situation made more or less salient the participants' relevant attitudes. In the basic condition, when attitudes were not salient, attitude-behavior correlations were .07, essentially zero. In the salient-attitude condition, participants were asked to “organize your thoughts and views on the affirmative action issue (e.g., Is it important to you that everyone is given equal opportunity in obtaining employment? Should women and minorities be actively recruited to help equalize employment ratios? …).” Under these circumstances, participants' prior attitudes predicted their verdict in the specific case, with a correlation of .58. Thus, the situation can emphasize or de-emphasize one's attitudes and hence their predictive power.

The situation can also emphasize different aspects of one's attitudes in different contexts. Merely thinking about one's attitude can make different aspects accessible in different circumstances, changing attitudes and undermining attitude-behavior correlations (Wilson et al., 1989). When people lack a well-defined set of reasons for their attitudes, they generate reasons on the spot, not necessarily reflecting their core beliefs, but instead what is verbalizable, accessible, plausible, and self-enhancing.

This points out, then, that the attitude object is ambiguous and shifts with context. For example (Smith, Fazio, & Cejka, 1996), one can categorize sunbathing as a beach activity or a cancer risk, and Lance Armstrong as a bike racer or a drug user, with different attitudes elicited in each case. Whichever category is more accessible in a given setting will guide attitudes and presumably behavior. Similarly (Sia, Lord, Blessum, Ratcliff, & Lepper, 1997), when different exemplars come to mind (i.e., for a black person, Oprah Winfrey and Bill Cosby versus a drug dealer and an armed robber), attitudes change and undermine attitude-behavior consistency. In line with these studies, people introspecting about reasons focus on whichever happen to be their most accessible thoughts, viewing them as applicable to their current attitude (Wilson, Hodges, & LaFleur, 1995). If different situations prime different aspects of attitudes, then attitudes and behavior will be unstable over time.

The setting matters in a broader way as well. The assumption that attitudes should predict behavior is a peculiarly Western one. In more interdependent, collective settings, using one's own personal views to guide social behavior might be seen as selfish and immature. In contrast, sensitivity—to the other's needs, to the group's goals, and to the situation's requirements—might be the more appropriate and valued response. In Japanese culture, for example, the belief in attitude-behavior consistency is less strong than in English-speaking countries, such as Australia (Kashima, Siegel, Tanaka, & Kashima, 1992). Thus, the larger situation, culture, matters as well.

IT DEPENDS ON THE PERSON

Some people more than others show attitude-behavior consistency. Recall individual differences in self-monitoring from the self chapter: High self-monitors are sensitive to and mold themselves to the demands of the situation, whereas low self-monitors are sensitive to and follow their inner selves as guides to behavior. It follows, then, that high self-monitors (those who monitor themselves to fit the situation) would show lower attitude-behavior correlations than low self-monitors. Kraus (1995) reports that four studies' average correlation for low self-monitors was a substantial .50, whereas high self-monitors' correlation was only .25.

A completely different kind of individual difference is need for cognition, mentioned earlier, the extent to which people seek and enjoy thinking hard. People high on need for cognition, as predicted, not only base their attitudes on issue-relevant thinking but also show higher attitude-behavior correlations: a whopping .87 between their attitudes toward presidential candidates before the fall election and their reported voting behavior eight weeks later; people low in need for cognition showed a correlation of .46 (Cacioppo, Petty, Kao, & Rodriguez, 1986).

Another way that attitude-behavior correlations depend on the person comes from researchers' habit of using college students as participants. David Sears (1986) has suggested that student attitudes are not as fixed as those of their elders and that the resulting instability may undermine attitude-behavior correlations. However, more recent evidence indicates that young adults and elderly adults both have less rigid attitudes, whereas middle-aged adults have strong attitudes, not susceptible to change, because they perceive their attitudes to be central, important, and well supported (Visser & Krosnick, 1998). Kraus (1995) reports a small but significant effect across studies, with an attitude-behavior correlation of .49 for nonstudents and .34 for students, but his nonstudent samples probably included more middle-aged than elderly adults, who were not analyzed separately.

IT DEPENDS ON THE PROCESS

Earlier, we described dual-process models of persuasion, contrasting more rapid, unconscious, automatic processes with slower, more conscious, deliberate processes of attitude change. Attitude-behavior linkages also depend on whether the process involved is relative more impulsive versus reflective (Strack & Deutsch, 2004). A reflective system operates on knowledge and values, to produce behavior, while the impulsive system operates through simpler and more direct associations and motivations. The systems operate in parallel, though the impulsive one is faster, it may respond first, all else being equal. In Fazio's (2001) related MODE approach (Motivation and Opportunity Determine Behavior), highly accessible attitudes generate spontaneous behavior, as mentioned in the earlier discussion of distinct types of attitudes predicting behavior. In both approaches, behavior can come from a rapid association between an attitude object, the attitude, and the relevant behavior, or behavior can come from more conscious deliberation, given capacity and motivation.

The unconscious operation of attitudes on behavior is the more surprising type of process. For example, people do not often consider their voting choices to be driven by automatic, unconscious processes, but they often are (Burdein, Lodge, & Taber, 2006). People often act on what “feels right,” as noted earlier with regard to perceptual fluency increasing pleasure (and stock purchases!). The broader principle is that people take feelings as information (Schwarz, 2004). Experiences that “go down easy” must be positive: fluent perception and accessible thoughts form the basis of judgments and behavior at the expense of more difficult knowledge retrieval. This may depend on the person: Prevention-focus participants prefer a vigilant nonverbal communicator, whereas promotion-focus people prefer an eager nonverbal communicator (Cesario & Higgins, 2008). Whichever style feels right has more influence.

People can be jolted out of this kind of impulsive, associative, go-with-your-gut process by experiencing difficulty in obtaining or using information (Alter, Oppenheimer, Epley, & Eyre, 2007), or presumably in applying it to behavior. Likewise, irrelevant affect has less influence not only at excessively high levels of effort, but also at excessively low levels of effort, when it barely registers (Albarracín & Kumkale, 2003).

Summary of Attitude-Behavior Research

Not all seemingly relevant attitudes predict seemingly relevant behavior. The level of specificity between attitudes and behavior has to match. Both attitude and behavior have to be measured carefully. Attitude-behavior correspondence depends on the attitude's mental accessibility and its overall strength. It also depends on the extent to which the situation or the person focuses on inner thoughts and feelings or on social norms as a guide to behavior, reflecting the core social motives of understanding versus belonging. Some people like to think, but not all introspection increases attitude-behavior correspondence; justifying reasons for one's attitude decidedly does not improve its predictive role. The attitude-behavior relation underscores the object-appraisal function of attitudes, akin to the core social motive of understanding. However, attitude-behavior consistency depends on culture as well. In the case of more collectivist, interdependent cultures, attitudes may more often serve value-expressive functions and therefore the core motive to belong. The process of generating the attitude and its behavior also determines the degree and type of linkage.

CHAPTER SUMMARY

Attitudes are evaluative responses to entities in the social world, and they have affective, cognitive, and behavioral correlates. Attitudes are most often measured by self-report, but other operational definitions supplement these. Attitudes all serve the object-appraisal function, equivalent to the core social motive of understanding. Some attitudes serve narrow self-interest functions, utilitarian goals. The other major attitude function, value-expression, resonates with the core social motive of belonging, either public or private identity with groups. Affective, cognitive, and behavioral processes relate to attitudes as primarily reflecting both understanding and belonging motives.

Attitudes often form through affective processes such as classical conditioning (repetition and association), instrumental conditioning (rewards and punishments), modeling (imitation), and vicarious reinforcement (observational learning). Another affect-based process of attitude formation is mere exposure, the increasing evaluation of entities frequently encountered. As noted, all these processes entail minimal cognitive activity of the systematic, thoughtful type that forms opinions, focusing instead on affect, such as pleasure and pain. Nevertheless, such primitive, affect-based attitudes serve the function of object appraisal.

Given an attitude, even if affect-based, object appraisal becomes richer and deepens understanding as people develop cognitions through a variety of processes, including consistency pressures and persuasive communication. Consistency theories are premised on people's tendency to fit together affect, cognition, and behavior relevant to a particular attitude. Accordingly, people seek and interpret information to support their attitudes. Cognitive consistency theories illustrate not only the core motive of understanding but also the core social motive of self-enhancement, seeing oneself as consistent and as not freely choosing actions with aversive consequences. The object-appraisal function of understanding is served by consistent, not arbitrary, attitudes. The attitude change that results from consistency pressures represents processes primarily internal to the person, described as self-inflicted attitude change.

Persuasion represents deliberate attempts to change the attitudes of another. The object-appraisal function, exemplifying the core social motive of understanding, especially represents the persuasive communication approaches. In analyzing persuasive communications for their effectiveness, researchers typically break down the process, as reflected in four major theories. The Yale communication and persuasion approach—both McGuire's stages of reception and yielding, as well as the larger communication and persuasion program of source, message, recipient, and context variables—set the scientific agenda for subsequent researchers, establishing persuasion as a topic, specifying variables, and establishing experimental approaches. However, the resulting research perhaps overemphasized deliberate, thoughtful persuasion processes, based on careful processing of message content. Other persuasion processes—considerably more superficial—matter as well. More recent models, namely the heuristic-systematic model and elaboration likelihood model, emphasize alternative routes to persuasion based on relatively superficial or in-depth processing. While the generalizations are few in this area of knowledge, the fine-grained understanding of effective persuasion techniques is well developed.

Having focused on affective processes in attitude formation and cognitive processes in attitude change, attitude research also concerns behavior. Just how much attitudes concern behavior proved contentious, as early researchers first assumed and then denied that attitudes predict behavior. Increased theoretical precision, consideration of omitted variables, and matching levels of attitude-behavior specificity all rescued attitude-behavior research from its predicament. Extending a theory of reasoned action, the theory of planned behavior specifies three predictors of behavioral intention: subjective beliefs, subjective norms, and perceived behavioral control; specific intentions, in turn, predict behavior. Moreover, attitudes predict behavior when the attitude itself is strong and accessible, when the situation encourages acting on one's attitudes (and not going along with social norms), and when the person's personality predisposes acting on attitudes. The attitude-behavior scandal is over, and the resolution emerges as one of social psychology's success stories.