It's better to light a candle than to curse the darkness

Introduction to

Beliefs, Attitudes, and Ideologies

Daryl J. Bem

Certain opinions seem to go together. People who support affirmative action also seem likely to advocate stronger gun control, to oppose capital punishment, and to hold a pro-choice position on abortion. On the surface these diverse opinions do not seem to follow from one another logically--there are even some implied inconsistencies among them. And yet, knowing that a person holds one of the opinions often enables us to predict correctly that he or she also holds the others. This is possible, in part, because the opinions all appear to follow from a common set of underlying beliefs, attitudes, and values--from an ideology.

Liberals are not the only ones with an underlying ideology. Those who oppose affirmative action and gun control laws often cite their belief in an underlying value of individual freedom as the basis for their opinions. Even those who disagree with such "conservative" opinions can appreciate the logic involved. But many such freedom-loving individuals also feel that a woman should not have the right to choose an abortion, that marijuana use should be more heavily penalized, and that homosexual behavior should be illegal. Here the logic is less than clear, yet these opinions, too, seem strangely predictable and ideologically consistent.

In this course we will explore the nature of such ideologies and attempt to understand how they get embedded into our individual psyches, our cultural discourses, and our social institutions. We shall see that the ideologies of a culture are typically invisible to the natives of that culture; they are, as computer types would say, "transparent to the user." Accordingly, a major goal of this course is to enable you to look at rather than through the lenses that filter, refract, and create your own current vision of reality.

Beliefs

Primitive Beliefs

Many of our beliefs are the product of direct experience. If you ask your friends why they believe oranges are round, they will reply that they have seen oranges, felt oranges, and that oranges are, indeed, round. That would seem to end the matter. But if you ask why they believe distant planets are round, the more sophisticated among them might be able to show how such a conclusion derives from physical principles and astronomical observations. If you press further by inquiring whence comes their knowledge of physical principles and astronomical observations, they might cite a professor or The New York Times. You can continue to probe: Why do they believe the professor or The New York Times?

What you will discover by such questioning--besides a decline in the number of your friends--is that every belief can be pushed back until it is seen to rest ultimately on a basic belief in the credibility of one's own sensory experience or on a basic belief in the credibility of some external authority. Other beliefs may derive from these basic beliefs, but the basic beliefs themselves are accepted as givens. Accordingly, I call them primitive beliefs.

Zero-Order Beliefs. Our most fundamental primitive beliefs are so taken for granted that we are apt not to notice that we hold them at all; we remain unaware of them until they are called to our attention or are brought into question by some bizarre circumstance in which they appear to be violated. For example, we believe that an object continues to exist even when we are not looking at it; we believe that objects remain the same size and shape as we move away from them even though their visual images change; and, more generally, we believe that our perceptual and conceptual worlds have a degree of orderliness and stability over time. Our faith in the validity of our sensory experience is the most important primitive belief of all.

These are among the first beliefs that children learn as they interact with the environment, and in a psychological sense, they are continuously validated by experience. As a result, we are usually unaware of the fact that alternatives to these beliefs could exist, and it is precisely for this reason that we remain unaware of the beliefs themselves. Only a very unparochial and intellectual fish is aware that its environment is wet. What else could it be? I will call primitive beliefs of this fundamental kind zero-order beliefs. They are the nonconscious axioms on which our other beliefs are built.

First-Order Beliefs. Because we implicitly hold these zero-order beliefs about the trustworthiness of our senses, particular beliefs that are based on direct experiences seem to carry their own justification. When we justify our belief in the roundness of oranges by citing our experiences with them, that does in fact usually end the matter. We do not run through a syllogistic argument of the form:

1st Premise: My senses tell me that oranges are round.

2nd Premise: My senses tell me true.

Conclusion: Therefore, oranges are round.

There is no conscious inferential process involved in going from the first premise to the conclusion because we take the second premise for granted: it is a zero-order belief. Accordingly, the first premise ("My senses tell me that oranges are round") is psychologically synonymous with the conclusion ("Oranges are round"). I call such conclusions first-order beliefs. Unlike zero-order beliefs, we are usually aware of our first-order beliefs because we can readily imagine alternatives to them (oranges could be pear shaped), but we are usually not aware of any inferential process by which we derive them from zero-order beliefs. Like zero-order beliefs, then, first-order beliefs are still appropriately called primitive beliefs, beliefs that demand no independent formal or empirical confirmation and that require no justification beyond a brief reference to direct experience.

Primitive Beliefs Based on External Authority. We not only experience our world directly, we are told about it as well. It is in this way that notions about such intangibles as God and germs first enter a child's system of beliefs. And to the child, such beliefs may seem as direct, as palpable, and as assuredly valid as any beliefs based on direct sensory encounter. When mommy says that not washing one's hands after pooping transmits dangerous germs, that is synonymous with the fact that not washing after pooping transmits germs. Such a belief is a primitive first-order belief for the child because the intervening premise, "Mommy says only true things," is nonconscious; the possibility that mommy sometimes says untrue things is not a conceivable alternative. First-order beliefs based on a zero-order belief in the credibility of an external authority, then, are functionally no different from first-order beliefs based on an axiomatic belief in the credibility of our senses. As sources of information, mommy and our senses are equally reliable. Our implicit faiths in them are zero-order beliefs.

This emphasis on the innocence of childhood should not obscure the fact that we all hold primitive beliefs. It is an epistemological and psychological necessity, not a flaw of intellect or a surplus of naïveté. We all share the fundamental zero-order beliefs about our senses, and most of us hold similar sorts of first-order beliefs. For example, we rarely question such beliefs as "this woman is my mother" or "I am a human being." Most of us even treat arbitrary social-linguistic conventions like "This is my left hand" and "Today is Tuesday" as if they were physical bits of knowledge handed down by some authority who "really knows." Finally, most religious and quasi-religious beliefs are first-order beliefs based on an unquestioned zero-order faith in some internal or external source of knowledge. The child who sings "Jesus loves me--this I know, / For the Bible tells me so" is actually being less evasive about the metaphysical--and hence nonconfirmable--nature of her belief than the founders of our nation were when they presumed to interpret reality for King George III: "We hold these truths to be self evident…"

Higher-Order Beliefs

Vertical Structure. Although we all hold primitive beliefs throughout our lives, we learn as we leave childhood behind to regard both our sensory experiences and external authorities as potentially fallible. We begin, in short, to insert an explicit and conscious premise about credibility between an authority's word and our belief.

The Surgeon General says that smoking causes cancer.
The Surgeon General is a trustworthy expert.
Therefore, smoking causes cancer.

We no longer treat the first premise as synonymous with the conclusion because the second premise is no longer a non-conscious zero-order belief. We are, for example, explicitly aware of the possibility that the Surgeon General might be in error. Accordingly, the conclusion "Smoking causes cancer" is not a primitive belief but a derived, or higher-order belief. It has a "vertical structure" of beliefs underneath it, beliefs that "generate" it as the product of quasi-logical inference.

We also derive higher-order beliefs by reasoning inductively from our experiences:

My aunt contracted cancer.
She died soon after.
Therefore, cancer can cause death.

And finally, we derive beliefs of still higher order by building on premises that are themselves conclusions in prior syllogisms. For example, the conclusions in the two syllogisms above can become premises in a new syllogism:

Smoking causes cancer.
Cancer can cause death.
Therefore, smokers die younger than nonsmokers.

Note that it is possible for two people to hold the same surface belief but to have different vertical structures underneath it. For example, the Surgeon General believes that smokers die younger on the average than non-smokers, but so also does the person who believes that

Smoking is a sin.
The wages of sin is death.
Therefore, smokers die younger than nonsmokers.

In this instance, the Surgeon General's belief is a higher-order belief based on a long chain of syllogistic reasoning, whereas for this person the same conclusion or surface belief is only a second-order belief, a belief derived from two first-order primitive beliefs.

Horizontal Structure. We might expect a higher-order belief to be quite vulnerable to disconfirmation because it would fall if any one of the underlying premises should be refuted. A higher-order belief would appear to be only as strong as its weakest link. This might be true if higher-order beliefs rested entirely on a single chain of inference, but many are bolstered by additional "horizontal" structures as well: a higher-order belief can be the conclusion to more than one chain of inference:

Those who derive their belief that "smokers die younger" from all three lines of inference will have their belief only partially weakened if one of the syllogisms contains flawed logic or one of the premises turns out to be false. Many of our higher-order beliefs rest not upon a single syllogistic pillar but upon many. They have broad horizontal as well as deep vertical structures.

Over time, the underlying structure of our higher-order beliefs can change. We might believe as we did before, but our reasons for believing have altered. The evidence on which we once based our trust of The New York Times may have been forgotten until now our devotion is a blind article of faith, a zero-order belief. Alternatively, additional support may have accumulated for beliefs that were once primitive beliefs or otherwise lacking in respectable justification.

Centrality. Some beliefs are more central to us than others in the sense that many of our other beliefs rest upon them. For example, the primitive zero-order belief in the general credibility of our senses is the most central belief of all; nearly all of our other beliefs rest upon it, and to lose our faith in it is, in the extreme case, to lose our sanity. Many of our religious beliefs and our beliefs about the nature of reality are similarly central; if they were to change, many of our other beliefs would also have to change as a consequence. In contrast, the belief that distant planets are round is probably not very central to most of us. If it turned out to be untrue, few of our other beliefs would have to change. This would be so even for those of us who have a deep vertical and wide horizontal structure behind the belief, that is, for whom the belief is derived from many different lines of reasoning. Formally stated, many syllogisms might lead up to our belief in round planets, but few syllogisms depart from it. A central belief enters as a premise into many belief syllogisms.

Logic versus Psycho-logic. Underlying this description of beliefs as elements of syllogisms is a model of the layperson as an informal or intuitive logician. But even if this were a valid model of how our belief systems are constructed, it does not imply that all our beliefs are necessarily valid or that we correctly follow the formal rules of deductive or inductive reasoning. Even when the reasoning is correct, conclusions to syllogisms can be wrong if any of the underlying premises is false.

Second, there are often inconsistencies between different higher-order beliefs even though the internal reasoning behind each separate belief is consistent within its own vertical structure. One line of reasoning leads to one conclusion; a second line leads to a contradictory conclusion. And finally, the reasoning itself is often faulty. For example, anxious parents often use the following syllogism to discourage marijuana use among their offspring:

Most heroin addicts started on marijuana.
Many young people are trying marijuana.
Therefore, many young people will become heroin addicts.

As many young people have suggested in rebuttal, most heroin addicts started on mother's milk. Therefore. . .

It should thus be clear that the syllogistic model of beliefs is a model not of formal logic but of psycho-logic. Moreover, it is designed primarily to characterize some of the apparent coherence of our belief systems; it is not necessarily a psychologically valid theory of how we perceive our beliefs or how our beliefs actually get formed in the first place. A number of contemporary social psychologists have constructed an alternative model, one which proposes that we should regard the layperson not so much as an intuitive logician as an intuitive empirical scientist.

The Layperson as an Intuitive Scientist

The model of the layperson as an intuitive scientist proposes that in attempting to understand our world, we face the same basic tasks as the formal scientist. For example, we need to observe or collect data ("My friend Chris asserts that women should have the right to obtain abortions"; "Lee Yamuri achieved the highest score on the math test"). We also attempt to detect covariation or correlation, to discern what goes with what ("Do most people who support the right to abortion also oppose capital punishment?" "On the average, do Asians seem to do better in math and science than non-Asians?"). [There are also several other tasks proposed in the complete model, but they are not pertinent to the present discussion (Nisbett & Ross, 1980).]

Our intuitive attempts to assess data and detect covariation are successful much of the time. But we also make a number of systematic errors and hold theories or expectations about our social worlds that interfere with arriving at valid beliefs. As we shall see, our theories often shape our perceptions of the data and distort our estimates of covariation.

Collecting Data

The first difficulty we face as informal scientists is collecting data in a systematic and unbiased way. When a survey researcher wants to estimate how many Americans support a woman's right to abortion, he or she takes great care to ensure that a random or representative sample of people are contacted so that the numbers of Catholics, Protestants, men, women, and so forth are interviewed in proportion to their percentage of the total population. But when we, as informal survey researchers, try to make this estimate intuitively, our major source of data is likely to be the people we know personally. Obviously this is not a representative sample of the population.

Another major source of data for us is the mass media, which also provide a nonrandom and nonrepresentative sample of data. For example, the media necessarily give more attention to a small number of antiabortion protesters publicly demonstrating at a medical clinic than they do to a larger number of people who silently support the clinic's abortion service. The media are not being biased here in the usual sense; they are simply reporting the news. But the data they give us are still not a reliable sample from which to estimate public opinion.

A survey researcher also keeps accurate records of the data. But in everyday life, we constantly accumulate information in our heads and then attempt to recall it from memory when we are later called upon to make some judgment. Thus, not only are the data we collect a biased sample in the first place, but the data we actually bring to bear on our social judgments are further biased by problems of selective recall.

Vividness. One of the factors that influences the information we notice and remember is its vividness. Research has shown that when both vivid and nonvivid information compete for our attention, our estimates and judgments are more influenced by the vivid information--even when the nonvivid information is more reliable and potentially more informative (Nisbett & Ross, 1980; Taylor & Thompson, 1982).

In one study, introductory psychology students who planned to major in psychology were given information about upper-level psychology courses and then asked to indicate which courses they planned to take. The subjects either heard two or three students make some informal remarks about each course in a face-to-face session or they saw a statistical summary of course evaluations made by past students in the courses. The subjects were more influenced in their choices by the face-to-face remarks than by the statistical summary--even when the summary was accompanied by written quotations of those same remarks. The vivid face-to-face information was more influential than the nonvivid written information even though the former was based on less complete and representative data (Borgida & Nisbett, 1977).

The vividness effect is a particular problem with information from the mass media. Even if reporters scrupulously gave equal coverage to both the vivid and nonvivid sides of an issue, our own information processing tendencies would supply the bias. Thus, even if a television newscast reports the results of a survey showing that a national majority supports abortion rights, we are still more likely to store and later recall the vivid pictures of the antiabortion protest when we intuitively try to estimate public opinion.

Schemata. Even if we could collect data in a systematic and unbiased way, our perceptions of the data can still be biased by our existing expectations and preconceptions--our theories--of what the data should look like. Whenever we perceive any object or event, we compare the incoming information with our memories of previous encounters with similar objects and events. Our memories of objects and events are not photograph-like reproductions of the original stimuli, but simplified reconstructions of our original perceptions. Such representations or memory structures are called schemata (or the singular, schema) and are the result of perceiving and thinking in terms of mental representations of classes of people, objects, events, or situations. The process of searching for the schema in memory that is most consistent with the incoming data is called schematic processing. Schemata and schematic processing permit us to organize and process an enormous amount of information with great efficiency. Instead of having to perceive and remember all the details of each new object or event, we can simply note that it is like one of our preexisting schemata and encode or remember only its most prominent features. Schematic processing typically occurs rapidly and automatically; usually we are not even aware that any processing of information is taking place at all.

For example, we have schemata for different kinds of people. When someone tells you that you are about to meet an extravert, you retrieve your extravert schema in anticipation of the coming encounter. The extravert schema consists of a set of interrelated traits such as sociability, warmth, and possibly loudness and impulsiveness. General person-schemata like these are sometimes called stereotypes. We also have schemata of particular persons, such as the president of the United States or our parents. We even have a schema about ourselves--a set of organized self-concepts stored in memory (Markus, 1977). When you see a job advertisement for a peer counselor, you can evaluate the match between your counselor schema and your self-schema to decide whether you should apply.

Research confirms that schemata help us to process information. For example, if people are explicitly instructed to remember as much information as they can about a person, they actually remember less than if they are simply told to try to form an impression of him or her (Hamilton, 1979). The instruction to form an impression induces the subjects to search for various person-relevant schemata that help them organize and recall material better. The self-schema also permits us to organize and process information efficiently. For example, people can recall a list of words better if they are told to decide whether each word describes themselves as they go through the list (Ganellen & Carver, 1985; Rogers, Kuiper, & Kirker, 1977).

Without schemata and schematic processing, we would simply be overwhelmed by the information that inundates us. We would be very poor information processors. But the price we pay for such efficiency is a bias in our perception of the data. Consider, for example, the impression you form of Jim from the following observations of his behavior.

Jim left the house to get some stationery. He walked out into the sun-filled street with two of his friends, basking in the sun as he walked. Jim entered the stationery store, which was full of people. Jim talked with an acquaintance while he waited to catch the clerk's eye. On his way out, he stopped to chat with a school friend who was just coming into the store. Leaving the store, he walked toward the school. On his way he met the girl to whom he had been introduced the night before. They talked for a short while, and then Jim left for school. After school Jim left the classroom alone. Leaving the school, he started on his long walk home. The street was brilliantly filled with sunshine. Jim walked down the street on the shady side. Coming down the street toward him, he saw the pretty girl whom he had met on the previous evening. Jim crossed the street and entered a candy store. The store was crowded with students, and he noticed a few familiar faces. Jim waited quietly until he caught the counterman's eye and then gave his order. Taking his drink, he sat down at a side table. When he had finished his drink he went home. (Luchins, 1957)

What impression do you have of Jim? Do you think of him as friendly and outgoing or shy and introverted? If you think of him as friendly, you agree with 78 percent of people who read this description. But examine the description closely; it is actually composed of two very different portraits. Up to the sentence that begins "After school, Jim left...," Jim is portrayed in several situations as fairly friendly. After that point, however, a nearly identical set of situations shows him to be much more of a loner. Whereas 95 percent of the people who are shown only the first half of the description rate Jim as friendly, only 3 percent of the people who are shown only the second half do so. Thus, in the combined description that you read, Jim's friendliness dominates the overall impression. But when individuals read the same description with the unfriendly half of the paragraph appearing first, only 18 percent rate Jim as friendly; his nonfriendliness leaves the major impression (see the table below). In general, the first information we receive has the greater impact on our overall impressions. This is known as the primacy effect.

Schematic Processing and the Primacy Effect: Once a schema of Jim has been established, later information is assimilated to it. (After Luchins, 1957)

The primacy effect has been found repeatedly in several different kinds of impression formation studies, including studies using real rather than hypothetical persons. For example, subjects who watched a male student attempt to solve a series of difficult multiple-choice problems were asked to assess his general ability (Jones, Rock, Shaver, Goethals, & Ward, 1968). Although the student always solved exactly 15 of the 30 problems correctly, he was judged more capable if the successes came mostly at the beginning of the series than if they came near the end. Moreover, when asked to recall how many problems the student had solved, subjects who had seen the 15 successes bunched at the beginning estimated an average of 21, whereas subjects who had seen the successes at the end estimated an average of 13.

Although several factors contribute to the primacy effect, it appears to be primarily a consequence of schematic processing. When we are first attempting to form our impressions of a person, we actively search in memory for the person schema or schemata that best match the incoming data. At some point we make a preliminary decision: This person is friendly (or some such judgment). We then assimilate any further information to that judgment and dismiss any discrepant information as not representative of the real person we have come to know. For example, when explicitly asked to reconcile the apparent contradictions in Jim's behavior, subjects sometimes say that Jim is really friendly but was probably tired by the end of the day (Luchins, 1957). Our theory of Jim, which has already been established, shapes our perception of all subsequent data about him. More generally, our subsequent perceptions become schema-driven and therefore relatively impervious to new data. There is truth to the conventional warning that first impressions are important.

Theories. Schemata are actually minitheories of everyday objects and events. But more elaborate theories also affect our perception of data. In an elegant demonstration of this, students who held strongly divergent beliefs about whether or not capital punishment acts as a deterrent against homicide read a summary of two purportedly authentic studies. One of the studies appeared to show that capital punishment was a deterrent, and the other appeared to show that it was not. The students also read a critique of each study that criticized its methodology. The results showed that students on each side of the issue found the study supporting their own position to be significantly more convincing and better conducted than the other study. The more unsettling result, however, was that after reading all the evidence on both sides, students were actually more convinced about the correctness of their initial position than they were at the beginning of the study (Lord, Ross, & Lepper, 1979). This implies that evidence introduced into public debate in the hope of resolving an issue--or at least moderating extreme views--will tend instead to polarize public opinion even further. Proponents of each side will pick and choose from the evidence so as to bolster their initial opinions (Nisbett & Ross, 1980).

Detecting Covariation

Detecting covariation, or correlation--discovering what goes with what--is a fundamental task in every science. Discovering that symptoms of an illness covary with the amount of environmental pollution or correlate with the presence of a virus is the first step toward a cure. And as intuitive scientists of human behavior, we perceive--or think we perceive--such correlations all the time ("People who are against capital punishment seem more likely to hold a pro-choice position on abortion"; "Asians seem to do better in math and science than non-Asians"). Our schemata of classes of persons--stereotypes--are actually minitheories of covariation: The stereotype of an extravert, a gay person, or a college professor is a theory of what particular traits or behaviors go with certain other traits or behaviors.

Research shows that we are not very accurate at detecting covariations. Once again, our theories mislead us. When our schemata or theories lead us to expect two things to covary, we overestimate the correlation between them, even seeing illusory correlations that do not exist. But when we do not have a theory that leads us to expect them to covary, we underestimate the correlation, even failing to detect a correlation that is strongly present in the data.

This was demonstrated by two researchers who were intrigued by the fact that clinical psychologists routinely report correlations between their clients' responses to projective tests and their personality characteristics, whereas research studies fail to find such correlations. For example, experienced clinicians have often reported that gay men are more likely than heterosexual men to see anal images, feminine clothing, and three other, similar kinds of images in Rorschach inkblots. Controlled studies, however, have not found any of these images to be correlated with a homosexual orientation (Chapman & Chapman, 1969). The researchers hypothesized that psychologists see these correlations because the reported images fit a popular stereotype, or schema, of male homosexuality. Several experiments have now confirmed this hypothesis.

In one, college students were asked to study a set of Rorschach cards. Each card contained the inkblot, a description of the image a client had reported seeing in it, and a statement of two personal characteristics that the client possessed. The images described included the five stereotyped images reported by clinical psychologists to be correlated with male homosexuality plus a number of other unrelated images (for example, images of food). The characteristics reported were either homosexuality ("has sexual feelings toward other men") or unrelated characteristics (for example, "feels sad and depressed much of the time"). The cards were carefully constructed so that no image was systematically associated with homosexuality.

After studying all the cards, subjects were asked to report if they had noticed "any general kind of thing that was seen most often by men" with the different characteristics. The results revealed that the students in this study--like experienced clinical psychologists--erroneously reported a correlation between the stereotyped images and homosexuality. They did not report any correlations between the nonstereotyped images and homosexuality.

The researchers then repeated the study, modifying the cards so that two of the nonstereotyped images (a monster image in one inkblot and an animal-human image in another) alwaysappeared with the characteristic of homosexuality--a perfect correlation. Despite this, subjects still reported seeing the nonexistent correlation with the stereotyped images more than twice as often as the perfect correlation with the nonstereotyped images.

As intuitive scientists, we are schema- or theory-driven. We see covariations our theories have prepared us to see and fail to see covariations our theories have not prepared us to see.

Persistence of Stereotypes. Perhaps it is not surprising that the inexperienced students in the study just described are misled by their stereotypes to see nonexistent correlations in the data. But why should this be true of experienced clinical psychologists? Why doesn't their daily contact with real data correct their mistaken perceptions of covariation? More generally, why do our stereotypes persist in the face of nonconfirming data?

We can illustrate some of the factors involved by representing the covariation task in a 2 x 2 table, as shown in the figure below. It displays some hypothetical data relevant to a popular stereotype similar to that explored in the Rorschach inkblot study: the stereotype that gay men display effeminate gestures. The table classifies a hypothetical sample of 1,100 men into the 4 cells of the table according to whether they have a homosexual or a heterosexual orientation and whether they do or do not display effeminate gestures.

The correct way to assess whether the two factors are correlated is to examine whether the proportionof homosexual men who display effeminate gestures (the left-hand column) is different from the proportionof heterosexual men who display effeminate gestures (the right-hand column). To do this, we must first add up the two cells in each column to find how many men with each kind of orientation there are in the sample. When we do this, we see that 10 out of 100, or 10 percent of the gay men display effeminate gestures and 100 out of 1,000, or 10 percent of the heterosexual men do so. In other words, there is no correlation in these data between sexual orientation and effeminategestures. It is important to note that to assess the correlation, we had to take into account all four cells of the table. Now consider what our intuitions would tell us if we encountered these data in daily life--where we do not have the data neatly laid out in front of us.

In our society, men with a homosexual orientation are in a minority, as are men who display effeminate gestures. When the two occur together (as in cell A, gay men with effeminate gestures), it is a particularly distinctive occurrence. This has two consequences. First, research has shown that people overestimate the frequency with which they have actually encountered such distinctive combinations (Hamilton & Gifford, 1976; Hamilton & Sherman, 1989). Second, even if we did not overestimate their frequency, we are still most likely to notice and to remember instances that fall into cell A and to remain oblivious to the instances that fall into the other cells of the table.

Part of the reason for this is that the relevant information is almost never available to us. In particular, we almost never have the opportunity to assess the frequency of cell C, the number of gay men who do not display effeminate gestures. Cell B also provides an inferential trap for some people. When they observe a man with effeminate gestures, they may simply assume he is gay even though they have no knowledge of his actual sexual orientation. He might belong to either cell A or cell B. But through circular reasoning, they illegitimately convert cell B disconfirmations of their stereotype into cell A confirmations. Note that it is the stereotype itself that leads them to make this inferential error--another instance of how our information processing is schema- or theory-driven.

But even if the data from cells other than cell A were available to us, it would not typically occur to us that we need to know this other information. We find it particularly difficult to take into account--or to understand why we need to take into account--cell D, the frequency of non-gay men who do not display effeminate gestures. Why is this difficult?

I noted earlier in this essay that we are more likely to notice and to remember vivid rather than nonvivid information. This is why cell A is noticed, remembered, and overestimated: gay men with effeminate gestures are distinctive and, hence, vivid. In contrast, there are not many events that are less vivid--and hence, less noticeable and less memorable--than events that do not occur. But this is precisely what cell D events are: nonevents. The non-gay man who does notdisplay effeminate gestures does not constitute a psychological event for us. It is difficult to notice or to appreciate the relevance of nonevents in daily life.

This difficulty was cleverly employed by Arthur Conan Doyle in his Sherlock Holmes story, "The Adventure of Silver Blaze," in which the famous detective is asked to discover who had stolen a prize race horse from its private stable during the night. Holmes draws the police inspector's attention to "the curious incident of the dog in the night-time." Puzzled, the inspector says, "The dog did nothing in the night-time." To which Holmes replies: "That was the curious incident." Holmes then deduces correctly that the horse was stolen by its own trainer--because the dog had not barked and, hence, must have known the intruder (Doyle, 1892/1981, p. 197).

The nonvividness of nonevents also leads the news media to promote and sustain stereotypes. When a gay man commits a murder--especially one with sexual overtones--both the sexual orientation and the murder are featured in the news story; when a heterosexual man commits a murder--even one with sexual overtones--sexual orientation is not mentioned. Thus cell A events are widely publicized--thereby fueling the stereotype--whereas cell B events are not seen as relevant to sexual orientation. And, of course, cell C and D events--men of any sexual orientation who do not commit murder--are not news. They are nonevents.

Self-Fulfilling Stereotypes. Our schemata influence not only our perceptions and inferential processes, but also our behavior and social interactions. And this, too, can sustain our stereotypes. In particular, our stereotypes can lead us to interact with those we stereotype in ways that cause them to fulfill our expectations. Thus, our stereotypes can become both self-perpetuating and self-fulfilling. Two studies illustrate this process.

In one, the investigators first noted that white job interviewers employed a less friendly manner when interviewing African-American applicants than when interviewing white applicants. They hypothesized that this could cause African-American applicants to come off less well in the interviews. To test this hypothesis, they trained interviewers how to reproduce both the less friendly and the more friendly interviewing styles. Applicants (all white) were then videotaped while being interviewed by an interviewer using one of these two styles. Judges who later viewed the tapes rated applicants who had been interviewed in a less friendly manner significantly lower on their interview performance than those who had been interviewed in the more friendly manner (Word, Zanna, & Cooper, 1974). The study thus confirmed the hypothesis that prejudiced individuals can interact in ways that actually evoke the stereotyped behaviors that sustain their prejudice.

A commonly held stereotype is that physically attractive individuals are more sociable, poised, and outgoing than less physically attractive individuals (Dion, Berscheid, & Walster, 1972). To find out if this stereotype could be self-fulfilling, researchers had unacquainted male and female college students talk to each other over the telephone for about 10 minutes. Before each conversation, the man was shown a photograph of either an attractive or an unattractive woman and told (falsely) that it was a photograph of his phone partner.

The conversations were recorded on two-track tapes, and analyses of the men's side of the conversations showed that those who believed they were talking to an attractive woman were friendlier, more outgoing, and more sociable than were men who believed they were talking to a less attractive woman. More interestingly, judges who listened to the woman's half of each conversation without hearing the male partner or knowing his belief about the woman's attractiveness rated women whose partners believed they were attractive as more sociable, poised, and humorous than women whose partners believed they were unattractive. The men's stereotype of physically attractive women became self-fulfilling in a 10-minute telephone conversation (Snyder, Tanke, & Berscheid, 1977).

Attitudes

Attitudes are likes and dislikes--favorable or unfavorable reactions to objects, people, situations, or any other aspects of the world, including abstract ideas and social policies: "I love oranges"; "I can't abide Republicans." But even though attitudes are feelings, they often depend in part on underlying "evaluative" beliefs about the attitude objects ("Oranges contain lots of vitamins"; "Republicans have no compassion for the poor"). Like other beliefs, evaluative beliefs can be primitive beliefs based on sensory experience ("Spinach tastes terrible") or on authority ("God is good"). They can also be higher-order beliefs, built in syllogistic fashion on prior beliefs:

Affirmative action programs will bring about racial equality.
Racial equality is desirable.
Therefore, affirmative action programs are desirable.

Similarly, the attitudes themselves--direct feelings of like or dislike--can be conceptualized as conclusions to syllogisms:

Spinach has a terrible taste.
I dislike terrible tastes.
Therefore, I dislike spinach.

This syllogism appears trivial because we so often like things that we evaluate positively and dislike things that we evaluate negatively that we typically do not distinguish between an evaluative belief and an attitude that follows directly from it. The first premise and the conclusion are treated as synonymous because the middle premise is so frequently true that it has become a nonconscious belief. "Terrible tastes" are disliked almost by definition.

But there are exceptions. Consider the following nonsyllogism:

Cigarettes taste terrible, cause cancer, make me cough, and offend others.
I dislike terrible tastes, cancer, coughing, and offending others.
But I still like cigarettes.

This logical nonsequitur could arise because the individual holds other evaluative beliefs that appear in other syllogisms (e.g., "Cigarettes relax me; I like being relaxed; therefore, I like cigarettes"). More likely, however, is that the attitude is much more strongly determined by noncognitive factors--certainly the case for addictive substances like tobacco. In short, evaluative beliefs about something may partially determine, but are not synonymous with, attitudes toward it.

Values

Just as an individual's higher-order beliefs can be traced back down through their syllogistic structures to their origins in first-order and zero-order primitive beliefs, so, too, higher-order attitudes are often found to rest upon basic values. For example, suppose that a woman who has a positive attitude toward money were asked to explain why. Her justification might translate into a syllogistic structure of the following form:

Money would allow me to retire.
I would like to retire.
Therefore, I like money.

When asked why she wants to retire:

Retirement would allow me to take piano lessons.
I would like to take piano lessons.
Therefore, I would like to retire.

When pushed further:

Music lessons would help me attain self-fulfillment.
Self-fulfillment is desirable.
Therefore, I would like to take music lessons.

Further questioning would reveal that the end of the syllogistic chain had been reached. That is, the evaluative belief "Self-fulfillment is desirable" (or, alternatively, the attitude "I would like self-fulfillment") would be seen by the individual as an end in itself, and, unlike money or retirement, not as a means to some other goal. No logical justification for wanting self-fulfillment would be seen as necessary or even possible; its desirability is self-evident. This is, of course, just a special case of what was defined earlier as a primitive belief; in this case, the primitive belief happens to be an evaluative belief or an attitude.

More concisely, then, a value can be defined as a primitive preference for or a positive attitude toward certain end-states of existence (like equality, salvation, self-fulfillment, or freedom) or certain broad modes of conduct (like courage, honesty, friendship, or chastity) (Rokeach, 1968). Values are ends, not means, and their desirability is either nonconsciously taken for granted--a zero-order belief--or seen as a direct derivation from one's experience or from some external authority--a first-order belief). To know whether a positive attitude or an evaluative belief is also a value for a particular individual, one must know the role it plays in his or her total belief system. One person's higher-order attitude can be another person's value. Money is a good example; it is a means to fulfilling other values for many but an end in itself for some.

Functions of Beliefs and Attitudes

Beliefs and attitudes serve a number of different psychological functions for us. Different people might hold the same belief or attitude for different reasons, and a person might hold a particular belief or attitude for more than one reason. The functions that beliefs and attitudes serve for the person also influence how consistent they are with one another and how easily they can be changed. Over the years social psychologists have identified and discussed a number of the functions that beliefs and attitudes might serve (Herek, 1986; Katz, 1960; Smith, Bruner, & White, 1956). Here are five of them.

Instrumental Function. Beliefs and attitudes that we hold for practical or utilitarian reasons are said to serve an instrumental function. They simply express specific instances of our general desire to obtain benefits or rewards and avoid punishment. For example, most Americans favor more government services but oppose higher taxes. As this example indicates, such attitudes are not necessarily consistent with one another. To change such attitudes, the person needs only to be convinced that some alternative would bring more benefits.

Knowledge Function. Beliefs and attitudes that help us to make sense of the world, that bring order to the diverse information we must assimilate in our daily lives, are said to serve a knowledge function. Such beliefs and attitudes are essentially schemata that enable us to organize and process diverse information efficiently without having to attend to its details. For example, until its recent demise, negative attitudes toward the Soviet Union helped many Americans to organize and interpret world events in terms of the cold war. The belief that Democrats just want to "tax and spend" or that Republicans care only for the wealthy provides a quick schematic way of interpreting and evaluating the proposals and candidates offered by the two parties. Like other schemata, such beliefs and attitudes often oversimplify reality and bias our perception of events.

Value-Expressive Function. Beliefs and attitudes that express our values or reflect our self-concepts are said to serve a value-expressive function. For example, a person might have positive attitudes toward gay people because of deeply held values about diversity, personal freedom, and tolerance; another person might have negative attitudes because of deeply held religious convictions that condemn homosexuality as a sin. Because value-expressive beliefs and attitudes derive from a person's underlying values or self-concept, they tend to be consistent with one another. Broad political values such as liberalism or conservatism often serve as a basis for value-expressive beliefs and attitudes. Such beliefs and attitudes do not change easily; the individual has to be convinced that alternative beliefs and attitudes would be more consistent with his or her underlying values or self-concept.

Ego-Defensive Function. Beliefs and attitudes that protect us from anxiety or from threats to our self-esteem are said to serve an ego-defensive function. The concept of ego defensiveness comes from Sigmund Freud's psychoanalytic theory. One of the mechanisms of ego defense Freud described is projection: The individual represses his or her own unacceptable impulses and then expresses hostile attitudes toward others who are perceived to possess those same impulses. For example, a person who is fearful of his or her own possible homosexual feelings might deny and repress such feelings and then display hostility toward gay people. (The term "homophobia" accurately describes anti-gay prejudice that serves such an ego-defensive function; it is less accurate when used more broadly to refer to all anti-gay attitudes.) In one study, students at a liberal California university were asked to write essays describing their attitudes toward lesbians and gay men. A content analysis of the essays revealed negative attitudes serving an ego-defensive function in about 35 percent of the essays (Herek, 1987).

The notion that negative attitudes toward minority groups can serve an ego-defensive function is called the scapegoat theory of prejudice, because the person's hostility often takes the form of blaming the groups for both personal and societal problems. The theory was tested in the late 1940s by a group of psychologists at the University of California at Berkeley. The research sought to discover whether psychoanalytic theory could explain the kind of anti-Semitism and fascist ideology that had emerged in Nazi Germany and whether one could identify individuals who would be particularly susceptible to such an ideology. The research, described in the book The Authoritarian Personality, has become a classic in social psychology (Adorno, Frenkel-Brunswik, Levinson, & Sanford, 1950).

Using belief and attitude questionnaires, the investigators first confirmed their hypothesis that individuals who were anti-Semitic were also likely to be prejudiced against many groups other than their own--called "outgroups." In interviews, such individuals recalled rigidly moralistic parental discipline, a hierarchical family structure, and an anxious concern about the family's socioeconomic status.

According to the investigators, such home environments produce individuals with "authoritarian personalities"--individuals who are submissive and obedient to those they consider their superiors (including authority figures), but contemptuous of and aggressive toward those they consider inferior. As the psychoanalytic theory of prejudice predicts, authoritarian individuals repress knowledge of their own undesirable characteristics, projecting them instead onto members of "inferior" outgroups.

Although the authoritarian personality study has been criticized for a number of shortcomings (Christie & Johoda, 1954), many of its original conclusions have withstood the test of continued research. In particular, there does appear to be an authoritarian personality type who seems particularly susceptible to a fascist ideology that has hostility toward outgroups at its core. More recent research does suggest, however, that prejudice and authoritarian attitudes may be acquired more directly in the home environment through the usual learning processes rather than through the more involved psychoanalytic processes described in the original research (Altemeyer, 1988).

Social Adjustment Function. Beliefs and attitudes that help us feel a part of a social community are said to serve a social adjustment function. People who hold the prescribed beliefs and attitudes of a particular church or political party because their friends, families, and neighbors do so provide a common example. The actual content of the beliefs and attitudes is less important than the social bonds they provide. To the extent that beliefs and attitudes serve primarily a social adjustment function, they are likely to change if the social norms change.

This was strikingly shown in the American South during the 1950s, when legalized racial segregation was being dismantled. Surveys showed that Americans in the South were generally opposed to desegregation and were more likely than Americans in the North to express negative attitudes toward African Americans. Some psychologists suggested that Americans in the South might be more authoritarian than Americans in other regions--that racial attitudes in the South were serving an ego-defensive function. But Thomas Pettigrew, a social psychologist who specializes in race relations, argued that racial attitudes in the South were being sustained primarily by simple conformity to the prevailing social norms of the region--a social adjustment function (Pettigrew, 1959).

Using the questionnaire developed for measuring authoritarianism, Pettigrew found that Southerners were no more authoritarian than Northerners (although authoritarian individuals in both regions were more prejudiced against African Americans than nonauthoritarian individuals). Moreover, Southerners who were prejudiced against African Americans were not necessarily prejudiced against other outgroups--which is contrary to what the theory of authoritarianism predicts. In fact, the South has historically been one of the least anti-Semitic regions in the United States, and one survey at the time showed Southern whites to be unfavorable toward African Americans but quite favorable toward Jews (Prothro, 1952). Also, veterans from the South--whose army experience had exposed them to different social norms--were considerably less prejudiced than nonveterans, even though veterans from both South and North were more authoritarian than nonveterans.

The subsequent history of desegregation confirmed Pettigrew's analysis. As desegregation progressed, surveys showed that attitudes toward a particular desegregation step tended to be unfavorable just before the change had been implemented but then became favorable soon afterwards (Pettigrew, 1959). Thus some communities had accepted the desegregation of public accommodations but were still opposed to school desegregation; other communities showed just the reverse pattern. In one study, it was estimated that about 40 percent of the sample had firm opinions favoring or opposing desegregation, but that the remaining 60 percent favored whatever the social norms happened to be at the time (Minard, 1952).

It is often said that one cannot legislate attitudes. In the literal sense this is obviously true. But legislation and judicial decrees can change public policies and practices, and these, in turn, frequently can change social norms. To the extent that a citizen's beliefs and attitudes are serving a social adjustment function, they, too, will change. Under these conditions, the quickest path to changing "hearts and minds" is to change behavior first by changing the social norms.

Social Norms and Reference Groups

The relatively rapid attitude changes seen in the American South during desegregation is but one example of the power that social norms can exercise over our beliefs and attitudes. Nearly every group to which we belong, from our immediate families to the culture as a whole, has an implicit or explicit set of beliefs, attitudes, and behaviors that are considered appropriate for its members. Any member of a group who strays from those norms risks isolation and social disapproval; in other words, groups regulate our beliefs and attitudes through the use of social rewards and punishments. But groups influence us in a second, more subtle way by providing us with a frame of reference, the lenses through which we look at the world and which define social reality for us. Any group that exercises either of these two kinds of influence is said to be one of our reference groups. We "refer" to them for deciding what to believe and how to feel and act.

An individual does not necessarily have to be a member of a reference group in order to be influenced by its norms. For example, lower-class individuals often use the middle class as a reference group. An aspiring athlete may use professional athletes as a reference group, adopting their views, their social norms--and perhaps their breakfast cereals.

Life would be simple if each of us identified with only one reference group. But we don't, so it isn't. A Jewish executive in a large corporation provides an example. Her ethnic reference group, the Jewish community, is characteristically liberal on most social and political issues and is heavily Democratic. But her business reference group is likely to be conservative, particularly on welfare and economic issues, and predominantly Republican. When issues or candidates are to be voted upon, such an individual often finds herself subjected to considerable "cross-pressures," both from her competing reference groups and, sometimes, from within herself.

But the reference group conflict that stands out above all others is that experienced by many young people between their families and their college or peer reference group. The most ambitious study of this conflict is Theodore Newcomb's classic Bennington Study--an examination of the political attitudes of the entire population of Bennington College. The dates of the study, 1935-1939, are a useful reminder that this is not a new phenomenon.

Today Bennington College is coed and tends to attract applicants who are aware of its politically liberal reputation. But in 1935, it was a women's college and most of the students came from politically conservative families--who, be it noted, could afford to send their daughters to an expensive college in the middle of history's worst economic depression. For example, over two-thirds of the parents of Bennington students were affiliated with the Republican party in the late 1930s.

At Bennington, these women encountered faculty members and older students who held a much more liberal perspective on world affairs (such as the Great Depression and the threat of a second World War) than their parents did. And as the women moved through their education at Bennington, they moved progressively further away from their parents' attitudes. For example, in the 1936 presidential campaign, 66 percent of their parents favored the Republican candidate, Landon, over the Democratic candidate, Roosevelt. So did about 62 percent of the Bennington freshmen. But only 43 percent of the sophomores favored Landon, and only 15 percent of the juniors and seniors did.

For most of the women, their increasing liberalism reflected a deliberate choice between the competing reference groups of college and family. Initially, many of them chose to go along with the college norms for pragmatic or non-intellectual reasons; their newly adopted attitudes served a social-adjustment function for them. Here are two examples:

All my life I've resented the protection of governesses and parents. At college I got away from that, or rather, I guess I should say, I changed it to wanting the intellectual approval of teachers and more advanced students. Then I found that you can't be reactionary and be intellectually respectable.

Becoming radical meant thinking for myself and, figuratively, thumbing my nose at my family. It also meant intellectual identification with the faculty and students that I most wanted to be like. (Newcomb, 1943, pp. 134, 131)

But as the women continued to mature, their adopted beliefs and attitudes began to become a genuine part of their ideological identities. In other words, their attitudes shifted from serving a purely social-adjustment function to serving a value-expressive function for them:

It didn't take me long to see that liberal attitudes had prestige value.... I became liberal at first because of its prestige value; I remain so because the problems around which my liberalism centers are important. What I want now is to be effective in solving problems.

Prestige and recognition have always meant everything to me.... But I've sweat blood in trying to be honest with myself, and the result is that I really know what I want my attitudes to be, and I see what their consequences will be in my own life. (Newcomb, 1943, pp. 136-137)

Did these changes in political attitudes become a part of an enduring ideological identity? In general, the answer is yes. Two follow-up studies of the Bennington women 25 and 50 years later found they had remained liberal. For example, in the 1984 presidential election, 73 percent of Bennington alumnae preferred the Democratic candidate Walter Mondale over the Republican candidate Ronald Reagan, compared with fewer than 26 percent of women of the same age and educational level. Moreover, about 60 percent of Bennington alumnae were politically active, most (66 percent) within the Democratic party (Alwin, Cohen, & Newcomb, 1991; Newcomb, Koening, Flacks, & Warwick, 1967).

Nevertheless, we never outgrow our need for identification with supporting reference groups. The political attitudes of Bennington women remained stable, in part, because they selected new reference groups after college--friends and husbands--who supported the attitudes they developed in college. Those who married more conservative men were more likely to be politically conservative in 1960. As Newcomb noted, we often select our reference groups because they share our attitudes, and then our reference groups, in turn, help to develop and to sustain our attitudes. The social-adjustment function of attitudes remains a potent force even when other functions are operative.

Consistency Revisited

I began this essay with the observation that certain opinions seem to go together. I implied that people do not simply subscribe to a random collection of beliefs and attitudes but have internally consistent belief systems and coherent ideologies. But psychologists and political scientists who study public opinion are divided in their views about the ideological coherence of public opinion on social and political issues (Kinder & Sears, 1985). One of those who believes the public to be ideologically innocent has said

As intellectuals and students of politics we are disposed by training and sensibility to take political ideas seriously.... We are therefore prone to forget that most people take them less seriously than we do, that they pay little attention to issues, rarely worry about the consistency of their opinions, and spend little or no time thinking about the values, presuppositions and implications that distinguish one political orientation from another. (McClosky, quoted by Abelson, 1968)

An example of such nonconsistency was revealed in a national survey taken by The New York Times and CBS News in the late 1970s. The survey showed that a majority of Americans said they disapprove of "most government-sponsored welfare programs." Yet over 80 percent said they supported: the government's "program providing financial assistance for children raised in low-income homes where one parent is missing" (Aid to Families with Dependent Children, a major welfare program); the government's program for "helping poor people buy food for their families at cheaper prices" (the federal food-stamp program); and the government's program for paying for health care for poor people (Medicaid). The strong support for these major welfare programs was similar among all types of people--rich and poor, liberal and conservative, Democrat and Republican.

An earlier national survey, designed specifically to probe this kind of inconsistency, found a similar contradiction between an ideological conservatism and an operational liberalism in attitudes toward welfare. One out of four Americans was conservative on questions concerning the general concept of welfare but simultaneously liberal on questions concerning specific welfare programs (Free & Cantril, 1967).

Despite these findings, we need to be cautious about accusing someone of being inconsistent, because his or her attitudes may simply be inconsistent with our own ideological framework; inconsistency may be in the eye of the beholder. For example, opposition to capital punishment is usually characterized as a liberal position, whereas opposition to legalized abortion is usually thought of as a conservative position. Yet there is a quite logical coherence to the views of a person who, being against all taking of life, opposes both capital punishment and legalized abortion. (Catholic clergy often hold this set of views, for example.) Another example is provided by libertarians, who are opposed to any government interference in our lives. They are conservative on economic issues--the free market, not the government, should control the economic system--and in their opposition to government enforced civil rights laws and affirmative-action programs. But they are liberal on personal social issues, believing, for example, that the government should not criminalize the use of marijuana or concern itself with our private sexual behavior. To libertarians, both liberals and conservatives are inconsistent.

Nevertheless, the evidence suggests that most citizens do not organize their beliefs and attitudes according to any kind of overall ideology; nonconsistency, if not inconsistency, seems more prevalent than consistency. This has led one psychologist to propose that many of our attitudes come packaged in opinion molecules. Each molecule is made up of (a) a belief, (b) an attitude, and (c) a perception of social support for the opinion. In other words, each opinion molecule contains a "fact," a "feeling," and a "following" (Abelson, 1968): "It's a fact that when my Uncle Charlie had back trouble, he was cured by a chiropractor [fact]"; "You know, I feel that chiropractors have been sneered at too much [feeling], and I'm not ashamed to say so because I know a lot of people who feel the same way [following]." Or, "Americans don't really want universal health insurance [following], and neither do I [feeling]. It would lead to socialized medicine [fact]."

Opinion molecules provide a final example of how beliefs and attitudes serve important social-adjustment functions. First, opinion molecules act as conversational units, giving us something coherent to say when a particular topic comes up in conversation. Second, they give a rational appearance to our unexamined agreement with friends and neighbors on social issues. But most important, they serve as badges of identification with our important social groups, reinforcing our sense of belonging to a social community. Thus, the "fact" and the "feeling" are less important ingredients of an opinion molecule than the "following."

If the general conclusion is that most citizens do not organize their beliefs and attitudes according to any kind of overall ideology, then whence comes the apparent coherence of beliefs and attitudes? How and why is it that "certain opinions seem to go together"?

The short answer is implied in this essay: We nonconsciously assimilate invisible, prepackaged ideologies from our reference groups--our families, our social groups, and the culture at large.

The long answer is what this course is all about.

References

Abelson, R. P. (1968). Computers, polls, and public opinion--Some puzzles and paradoxes. Transaction, 5, 20&endash;27.

Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian personality. New York: Harper.

Altemeyer, B. (1988). Enemies of freedom: Understanding right-wing authoritarianism. San Francisco: Jossey-Bass.

Alwin, D. F., Cohen, R. L., & Newcomb, T. M. (1991). Personality and social change: Attitude persistence and changes over the lifespan. Madison: University of Wisconsin Press.

Borgida, E., & Nisbett, R. E. (1977). The differential impact of abstract vs. concrete information on decisions. Journal of Applied Social Psychology, 7, 258-271.

Chapman, L. J., & Chapman, J. P. (1969). Illusory correlation as an obstacle to the use of valid psychodiagnostic signs. Journal of Abnormal Psychology, 74, 271-280.

Christie, R., & Johoda, M. (Eds.). (1954). Studies in the scope and method of "The authoritarian personality." New York: Free Press.

Dion, K. K., Berscheid, E., & Walster, E. (1972). What is beautiful is good. Journal of Personality and Social Psychology, 24, 285-290.

Doyle, A. C. (1892/1981). The original illustrated Sherlock Holmes. Secaucus, NJ: Castle Books. (Originally published in America by Harper & Bros. in McClure's Magazine ,1893).

Free, L. A., & Cantril, H. (1967). The political beliefs of Americans. New Brunswick, NJ: Rutgers University Press.

Ganellen, R. J., & Carver, C. S. (1985). Why does self-reference promote incidental encoding? Journal of Personality and Social Psychology, 21, 284-300.

Hamilton, D. L., & Gifford, R. K. (1976). Illusory correlation in interpersonal perception: A cognitive basis of stereotypic judgments. Journal of Experimental Social Psychology, 12, 392&endash;407.

Hamilton, D. L., & Sherman, S. J. (1989). Illusory correlations: Implications for stereotype theory and research. In D. Bar-Tal, C. F. Graumann, A. W. Kruglanski, & W. Stroebe (Eds.), Stereotypes and prejudice: Changing conceptions. New York: Springer-Verlag.

Herek, G. M. (1986). The instrumentality of attitudes: Toward a neofunctional theory. Journal of Social Issues, 42, 99-114.

Herek, G. M. (1987). Can functions be measured? A new perspective on the functional approach to attitudes. Social Psychology Quarterly, 50, 285-303.

Jones, E. E., Rock, L., Shaver, K. G., Goethals, G. R., & Ward, L. M. (1968). Pattern of performance and ability attribution: An unexpected primacy effect. Journal of Personality and Social Psychology, 9, 317-340.

Katz, D. (1960). The functional approach to the study of attitudes. Public Opinion Quarterly, 24, 163-204.

Kinder, D. R., & Sears, D. O. (1985). Public opinion and political action. In G. Lindzey, & E. Aronson (Eds.), The handbook of social psychology (3rd ed., Vol. 2, pp. 659&endash;742). New York: Random House.

Lord, C. G., Ross, L., & Lepper, M. R. (1979). Biased assimilation and attitude polarization: The effects of prior theories on subsequently considered evidence. Journal of Personality and Social Psychology, 37, 2098-2109.

Luchins, A. (1957). Primacy-recency in impression formation. In C. I. Hovland (Ed.), The order of presentation in persuasion. New Haven: Yale University Press.

Markus, H. (1977). Self-schemata and processing information about the self. Journal of Personality and Social Psychology, 35, 63-78.

Minard, R. D. (1952). Race relations in the Pocahontas coal field. Journal of Social Issues, 8, 29&endash;44.

Newcomb, T. M. (1943). Personality and social change. New York: Dryden Press.

Newcomb, T. M., Koenig, K. E., Flacks, R., & Warwick, D. P. (1967). Persistence and change: Bennington College and its students after twenty-five years. New York: Wiley.

Nisbett, R. E., & Ross, L. (1980). Human inference: Strategies and shortcomings of social judgment. Englewood Cliffs, NJ: Prentice-Hall.

Pettigrew, T. F. (1959). Regional differences in anti-Negro prejudice. Journal of Abnormal and Social Psychology, 59, 28&endash;36.

Prothro, E. T. (1952). Ethnocentrism and anti-Negro attitudes in the deep South. Journal of Abnormal and Social Psychology, 47, 105&endash;108.

Rogers, T. B., Kuiper, N. A., & Kirker, W. S. (1977). Self-reference and the encoding of personal information. Journal of Personality and Social Psychology, 35, 677-688.

Rokeach, M. (1968). Beliefs, attitudes, and values. San Francisco: Jossey-Bass.

Smith, M. B., Bruner, J. S., & White, R. W. (1956). Opinions and personality. New York: Wiley.

Snyder, M. L., Tanke, E. D., & Berscheid, E. (1977). Social perception and interpersonal behavior: On the self-fulfilling nature of social stereotypes. Journal of Personality and Social Psychology, 35, 656-666.

Taylor, S. E., & Thompson, S. C. (1982). Stalking the elusive "vividness" effect. Psychological Review, 89, 155-181.

Word, C. O., Zanna, M. P., & Cooper, J. (1974). The nonverbal mediation of self-fulfilling prophecies in interracial interaction. Journal of Experimental Social Psychology, 10, 109-120.