The Persistence of Stereotypic Beliefs: a Cognitive View

Vivian E. Peeters, University of California, Los Angeles
ABSTRACT - The persistence of stereotypic beliefs is examined from a cognitive viewpoint. Experimental evidence from the areas of illusory correlation. memory processes, belief perseverance, and hypothesis testing is examined in relation to the stereotypic belief persistence effect. The biasing impact of persisting stereotypic beliefs on information processing and subsequent judgments is discussed. Suggestions for future research and implications for the study of consumer behavior are presented.
[ to cite ]:
Vivian E. Peeters (1983) ,"The Persistence of Stereotypic Beliefs: a Cognitive View", in NA - Advances in Consumer Research Volume 10, eds. Richard P. Bagozzi and Alice M. Tybout, Ann Abor, MI : Association for Consumer Research, Pages: 454-458.

Advances in Consumer Research Volume 10, 1983      Pages 454-458


Vivian E. Peeters, University of California, Los Angeles

[The author thanks Carol Scott, Jim Bettman, Hal Kassarjian, and Rich Lutz for very helpful comments on an earlier draft.]

[Vivian Peeters is a doctoral student in Marketing at UCLA.]


The persistence of stereotypic beliefs is examined from a cognitive viewpoint. Experimental evidence from the areas of illusory correlation. memory processes, belief perseverance, and hypothesis testing is examined in relation to the stereotypic belief persistence effect. The biasing impact of persisting stereotypic beliefs on information processing and subsequent judgments is discussed. Suggestions for future research and implications for the study of consumer behavior are presented.


That people tend to cling to their beliefs is clear from both observation of everyday life and much social psychological research. For example, Ehrlich (1973) analyzed Americans' stereotypic beliefs about Russians, Japanese, and Germans over a 24-year period. Although some changes did occur, the perceptions of these groups remained highly stable.

Belief persistence and belief change have been of interest to consumer behavior researchers as well. Marketers traditionally have been concerned with maintaining favorable consumer beliefs or opinions and changing negative ones. More recently, marketers have had to contend with the effects of persistent beliefs developed from rumors. Procter & Gamble's ongoing battle with the rumor of its connection with Satanism or aspirin manufacturers' problems with the purported, but unproved, link between children's aspirin and Reye's syndrome are but two recent examples. In previous years, McDonald's was said to be grinding worms into its hamburgers, while K-Mart was charged with having snakes in its clothing. Such rumors have disastrous effects on sales, profits, corporate image, and consumer confidence (Moreland 1982). Marketers need to know how to combat unfounded rumors and negative information effectively and expediently.

Tybout, Calder, and Sternthal (1981) have demonstrated that two strategies based on the mechanisms of storage and retrieval of information are effective for counteracting rumors. An examination of other research related to belief persistence may provide further insights and suggest new approaches to belief change. This paper reviews cognitive research streams investigating the persistence of stereotypic beliefs, including those on illusory correlation, encoding and retrieval biases, belief perseverance in the presence of discredited information, and hypothesis-testing. New ways or inducing belief change which can be tested are derived and discussed.


Illusory correlation is the perception of a covariation between two events which, in reality, does not exist. Thus, even when presented with data regarding two unrelated and independent events, subjects may perceive that the two events are somehow associated and contingent upon one another.

Basic Findings

Various studies provide evidence for the illusory correlation phenomenon. For example, Smedslund (1963) presented nursing students with a series of cards, each card supposedly providing information about a different patient. Letters presented on the card indicated the presence or absence of a particular symptom in a patient who had or had not been diagnosed as having a particular disease. The subject's task was to determine whether or not there was a relationship between symptom and disease. Thus, the general correlational design can be illustrated as a frequency table representing the relationship between two variables, as follows:


Using a similar paradigm, Ward and Jenkins (1965) presented subjects with a series of items, each of which indicated that clouds had or had not been seeded and that subsequently it had or had not rained. The subjects' task was to judge the degree to which cloud-seeding was related to rainfall.

The results of both studies are similar and can be summarized by two points. First, subjects' ability to accurately judge the degree of relationship between two variables was quite low. Second, subjects' judgments of the degree of association between variables were primarily a function of the absolute frequency of instances of a particular X-Y pair, specifically the X-Present, Y-Present pair. Thus, the frequency or joint occurrences of the two events being considered (e.g. cloud-seeding and rain or the presence of a symptom and the disease) was the primary heuristic used to judge the relationship between the two variables. Other X-Y pairings had little impact on the subjects' judgment.

Effect of Associative Strength and Distinctiveness

Other work in the area of illusory correlation has provided findings which can be applied to the issue of stereotypic belief persistence. Chapman (1967), using word pairs, found that subjects significantly overestimated the frequency of occurrence of word pairs that were either 1) high in associative strength (knife-fork); or 2) distinctive because they contained atypically long words (three-syllable words in an otherwise one-syllable word list). Thus, particular stimuli pairing led to systematically biased judgments. Chapman concluded that both associative connections between stimulus events and the joint occurrence of distinctive stimulus events can produce illusory correlation.

Further evidence for the effect of distinctiveness on illusory correlation is presented by Hamilton and Gifford (1976). Based on Chapman's (1967) results, these experimenters hypothesized that observers would overestimate the frequency with which members of minority groups commit undesirable acts. To test this hypothesis, subjects read 39 sentences about different people. Each person, identified by first name and membership status in one of two groups, was described as having performed some behavior. Majority and minority groups were created by having 26 of the 39 sentences describe persons belonging to Group A and 13 of the sentences describe members of Group B. In addition, a majority of the sentences (27) described desirable behaviors, while a minority (12) described undesirable behaviors. This manipulation rendered both membership in Group B and undesirable behaviors distinctive by their relative infrequency within the context of the total stimulus set. Objectively, there was no relationship between group membership and the desirability of behaviors. In fact, the contingency table for the data is as follows:


As can be seen, the frequency of desirable and undesirable behaviors performed by members of Group A is exactly double that for Group B, although the relationship between the two variables is identical. Hamilton and Gifford predicted that the frequency of occurrence of the four instances of undesirable behavior performed by members of Group 8 (which represent the co-occurrence of distinctive events) would be overestimated by the subjects. This is precisely what happened, indicating the presence of illusory correlation. The subjects perceived a relationship between group- membership and behavior that, in fact, did not exist.

Another interesting aspect of this study is that subjects rated Group A members more favorably than Group B members on a series of 20 trait scales. This finding logically follows from the overestimation of the frequency with which Group B members committed undesirable behaviors.

This study shows then that an illusory correlation based on a cognitive bias in perceivers' information processing of two events results in the differential perception of two groups of stimulus persons. Thus, there is support for a cognitive mechanism operating in the stereotyping process.

Effect of Consistent and Inconsistent Information

The presentation of information either consistent or inconsistent with existing beliefs also seems to affect individuals' judgments of the association between two events. In order to test the effect of prior beliefs on the processing of various kinds of information, Hamilton and Rose (1980) presented subjects with sentences in which the members of different occupational groups were described by pairs of trait adjectives. The trait adjectives were either consistent with or unrelated to stereotypic beliefs. For example, trait adjectives consistent with a doctor stereotype were "thoughtful" and "wealthy," while those consistent with a stewardess stereotype were "attractive" and "comforting." Subjects were asked to estimate how frequently each of the trait adjectives had described members of each of the occupational groups. It was predicted that subjects would estimate that information congruent with stereotypic expectations had occurred more often than equally frequent information unrelated to or incongruent with the stereotype. The results upheld this hypothesis. Thus, the subjects' stereotypic beliefs were sustained, and in effect received confirmation, in the absence of substantiating evidence.

A related issue of how information which is disconfirming or incongruent with stereotypic expectancies is processed was also examined by Hamilton and Rose (1980), using a similar experimental design to the one described above. The major difference, however, was that the traits used were either incongruent with or unrelated to the occupational stereotypes. The results show that when there was no actual correlation between the traits and occupations, subjects estimated incongruent traits as occurring significantly less frequently than neutral traits. These findings again indicate that subjects' judgments were biased in the direction of maintaining their stereotypic beliefs. On the other hand, when the traits were strongly correlated with the occupational groups, this bias was not present. This result seems to suggest that when the evidence is overwhelmingly discrepant with the person's stereotypic expectancies, the initial-bias may be to some extent over-come. Thus, we have at least some encouragement for the possibility of attenuating a persisting stereotypic bias.


Given that illusory correlation exists and that distinctive stimuli are more attended to in the environment, these findings suggest that information is distorted in cognitive processing. Various investigators have examined the effects of memory processes on stereotypic beliefs and have found evidence for the occurrence of biases in both encoding and retrieval.

Evidence for Biased Encoding

Evidence for biased encoding is found in a study by Rothbart, Evans, and Fulero (1979). In this study, memory for information that confirmed or disconfirmed stereotypic expectancies was examined. Subjects were presented with a set of 50 behavior descriptions, where each behavior was associated with the name of one member of a group of 50 men. The 50 items consisted of 17 intelligent, 17 friendly, 3 nonintelligent, 3 unfriendly, and 10 unrelated behaviors. Half of the subjects were told that the group about which they were reading was intellectual; the other half was told that the group was friendly. In addition, half of the subjects were given the intellectual/friendly expectancy prior to presentation of the behaviors; the other half received the expectancy afterward. Subjects exhibited superior recall and higher frequency estimates for behaviors confirming the stereotype when the expectancy was induced prior to presentation of the behaviors; no such effects were observed for expectancies induced after the behaviors were presented. Rothbart et al. concluded that subjects' superior recall of expectancy-confirming events may account for the self-perpetuating character of social stereotypes, and that memory effects were not due only to the selective retrieval of behavioral events. Rather, the biasing effects of expectancies on memory may have important implications in the persistence of stereotypic beliefs.

Evidence for Biased Retrieval

Other studies have looked at the role of biased retrieval on stereotyping. For example, Snyder and Uranowitz (1978) found systematic retrospective distortions of past events as a result of one's current beliefs about another individual. In this study, subjects read a narrative about the life of a woman. Either immediately after reading the case history or one week later, some of the subjects learned that she was currently living a lesbian life-style; others learned that she was currently living a heterosexual life-style; the remaining subjects learned nothing about her life-style. The impact of this new information on recognition memory was then measured. The results show that subjects reconstructed the woman's past in ways that supported their current stereotype interpretations of her sexual life-style. The timing manipulation had no effect on the subjects' memory. These results demonstrate that current beliefs can exert powerful effects on attempts to remember the past and lend further support to the persisting nature of stereotypic beliefs.

Evidence for Biased Encoding and Retrieval

From the results of the studies cited above, it seems reasonable to assume that both biased encoding and biased retrieval processes are operating during stereotyping. Experimental evidence provided by Cohen (1981) supports this assumption. In two experiments, the influence of stereotypic knowledge on social perception in a realistic setting was examined. In Experiment 1, subjects viewed a videotape of a target woman identified either as a waitress or a librarian. Subjects more accurately remembered (via recognition) features of the woman that were consistent with their prototype of a waitress (librarian) than features that were inconsistent. For example, subjects were more likely to remember that the librarian in the videotape was wearing glasses and listening to classical music than that she was drinking beer and watching television. In Experiment 2, subjects learned the occupational information either before or after watching the tape. As in the first experiment, the prototype-consistency effect was replicated, even for the group who received the occupational label after viewing the videotape. However, knowing the target's occupation while watching the videotape led to increased accuracy for both consistent and inconsistent information.

Thus, this study and the other studies cited provide evidence for both biased encoding and retrieval in the stereotype process. From the literature reviewed up to this point, we can see strong cognitive foundations which support the persistence of stereotypic beliefs. A logical question to ask now is, What happens when the cognitive foundations are shaken? Will stereotypic beliefs persist when their basis is unfounded?


Answers to the above questions are available. Various studies have shown that stereotypes persist even when their initial evidential data base has been totally discredited. Using a debriefing paradigm, Anderson, Lepper, and Ross (1980) presented subjects with two case studies suggestive of either a positive or a negative relationship between risk-taking and success as a fire-fighter. Some subjects were asked to write an explanation for the relationship, while others were not. At the end of the experiment, subjects were thoroughly debriefed concerning the fictitious nature of the initial case studies. The results indicated that the debriefing, or total discrediting of the evidence on which subjects' initial theories had been based, had only a minimal impact on subjects' beliefs concerning the relationship between risk preference and fire-fighting ability. This result was true both for subjects who wrote an explanation and for those who did not. Debriefed subjects who had been initially exposed to data indicative of a positive relationship continued to believe that a positive relationship existed. Similarly, debriefed subjects in the negative relationship condition continued to believe in a negative relationship. Thus, this study provides strong evidence for unwarranted theory perseverance. The results are especially noteworthy when one considers that such strong belief persistence was based on a weak evidential base consisting of only two case histories.

In a similar study of the perseverance effect, Ross, Lepper, Strack, and Steinmetz (1977) had subjects read a description of a young woman. After reading it, the subjects were told that she had either committed suicide or that she had made substantial financial contributions to the Peace Corps. Subjects were then asked to write an explanation of how this event came about. After they had written their explanation, subjects were told that the event they had written about had not actually occurred. They were then asked to estimate the likelihood of several events occurring in the woman's life, including the event they had written about. Subjects consistently overestimated the probability that the event they had written about would occur as compared to other events. It would appear that the writing exercise caused the subjects to become aware of the ways in which the event could have come about and the factors in the woman's -background had would lead to the event's occurrence.

Both of these studies utilized "outcome" debriefing procedures. That is, subjects were merely told that they had received false information and that they had been randomly assigned to experimental conditions. Another type of approach which has also been used is "process" debriefing, where subjects are presented with both the standard outcome debriefing conditions and an extensive discussion of the perseverance phenomenon, the processes that might contribute to it, and the potential personal relevance and costs of erroneous impression perseverance. This process debriefing approach was found to have partial success in a study by Ross, Lepper, and Hubbard (1975). In that study, actors who had personally received the process debriefing showed little belief persistence. However, observers of these same actors and of their process debriefing continued to exhibit belief maintenance.

Thus, all of these studies indicate the persevering nature of beliefs and the difficulty people have in separating objective information from the inferential structure surrounding it.


By now we can clearly see that stereotypic beliefs are extremely persistent and resistant to change. But what if one deliberately cries to test some hypothesis in order to determine its veridicality?

As Snyder (1981) so aptly points out, "Seek and ye shall find." Put yet another way, "What you expect is what you gee." In a series of experiments, Snyder examined how people actively test hypotheses about other people. Specifically, subjects were provided with information as to the introvertedness or extravertedness of target persons. The subjects' task was to choose a series of questions to ask their targets in order to test an introversion or extraversion hypothesis. Snyder's results can be summarized as follows:

1. Individuals planned to test their hypothesis by soliciting behavioral evidence whose presence would tend to confirm their hypothesis.

2. The origin of the hypothesis (i.e. whether from a credible source or not) had no effect on subjects' use of confirmatory hypothesis-testing strategies.

3. The certainty (i.e. statistical likelihood) of the hypothesis had no effect on subjects' preference for confirmatory strategies.

4. Incentives for accuracy had no effect on the propensity to solicit hypothesis-confirming evidence from the targets.

5. Even when faced with competing hypotheses to test (i.e. whether the target is an introvert or extravert), individuals continued to prefer a confirmatory hypothesis-testing strategy, usually from a single domain (either introvert or extravert).

6. When given both confirming and disconfirming attributes, subjects continued to use confirmatory hypothesis-testing strategies.

7. The confirmatory hypothesis-testing procedures channeled actual social interaction between the hypothesis-testers and targets such that the targets provided actual behavior which confirmed the hypothesis. That is, targets who were being "tested" for introversion actually behaved in an introverted manner (e.g. were shy and reserved).

Thus, Snyder's results show that people have a very strong predilection for confirming that in which they already believe. The only time that Snyder's subjects did not engage in a confirmatory strategy was when they were presented with no hypothesis to test. This situation, by definition, is not applicable to our examination of the persistence of already-held stereotypic beliefs.

These findings indicate that hypothesis-confirming evidence will be over-represented and hypothesis-disconfirming evidence will be under-represented even after an attempt to "objectively" gather information about one's beliefs. Thus, attempts by individuals to monitor and test the accuracy of their hypotheses may turn hypotheses into self-confirming hypotheses.


Summary of the Findings

From this selected view of the literature, one can see that stereotypes do exist and do persist. Illusory correlation plays an important role in belief persistence, such that confirmatory instances of covariation between two objectively independent variables are perceived as occurring together more frequently than they actually do. Depending on the associative strength and distinctiveness of the two variables, the perceived contingent relationship between them may be exaggerated, since individuals are more likely to attend to distinctive stimuli in the environment. The literature also shows that people do not seem to require a sizable evidential data base on which to form their beliefs. Further, beliefs are highly resistant to change even when a somewhat questionable belief base is totally discredited, as in an experimental debriefing paradigm. Evidence for the role of memory processes in the belief persistence effect is also available. Both encoding and retrieval processes may operate to bias perception and memory for events. Finally, when testing hypotheses against reality, subjects tend to seek, and actually do obtain, information which confirms their original hypotheses. Thus, even attempts to test stereotypic beliefs can lead to unwarranted support for and strengthening of those beliefs.

Suggestions for Research

The task for future researchers is to use the knowledge of the cognitive mechanisms underlying, belief persistence to bring about belief change. The research suggests testing some of the following strategies.

Distinctive stimuli and illusory correlation. It seems that one effective way of bringing about belief change would be to manipulate the distinctiveness of both the object and attribute of interest and then to foster some new association between the two. Such a manipulation could be conducted in an experimental setting to deter mine if a new illusory correlation can be established among subjects and how effective that correlation is in displacing or attenuating a previously formed correlation.

In the case of a manufacturer, this manipulation could be achieved through advertising. Thus, Procter & Gamble might consider making distinctive some very positive product or company attribute, in order to weaken consumers' perceptions of the P&G-Satanism, link. Since P&G (the object) may be considered distinctive already by virtue of its size and dominance in the consumer marketplace, little strengthening of its distinctiveness would probably--be necessary. Only the selection of some positive and distinctive attribute for promotion might be needed for belief change.

Inconsistent information. The Hamilton and Rose (1983) study shows that information which is highly discrepant with a person's prior beliefs may be effective in over coming stereotypic expectancies. The implication is that refutational advertising might be effective if (and only if) it is perceived as being highly discrepant with existing beliefs. One technique for achieving high discrepancy is by increasing the actual correlation between the stimulus object and the incongruent information. This method suggests a large dosage of refutational advertising. However, several related issues still need to be examined before this technique can be recommended with confidence, especially in view of the ineffectiveness of refutational advertising in the Tybout, Calder, and Sternthal (1981) study and in the real world. The impact of the following factors should be assessed: the amount of inconsistent information, the degree to which the information is incongruent with existing beliefs, and the strength of preexisting beliefs. The results of such an examination may help to explain why refutational advertising was ineffective in the Tybout, Calder, and Sternthal study. Perhaps the inconsistent information in that study was not presented long enough or was of insufficient strength to effect belief change.

Debriefing paradigm. Another possibility for belief change which should be tested in the consumer behavior area is that of using an outcome debriefing paradigm which employs a concrete, vivid discrediting of the evidential base. If one examines the techniques used in the debriefing studies, one sees abstract and pallid factual information being presented to subjects in an effort to offset rich and colorful anecdotal data presented in the original case histories. Similarly in the real world case of McDonald's, factual data such as "100% pure beef" and "Red worms cost between $5 and $8 a pound" was used (ineffectively) to counteract the vivid "worm" rumors. Perhaps refutation via a counter-case equally vivid to the original, evidential cases might prove successful in the outcome debriefing paradigm. The findings of Borgida and Nisbett (1977) lend support to this kind of approach.

One might also consider further the process debriefing paradigm, which has been found to be somewhat successful in discrediting prior beliefs. It would be interesting to test this paradigm in a consumer behavior setting. However, the possibility of an extensive debriefing seems difficult to achieve in an everyday situation, where such debriefing would probably be accomplished through mass communication. The cost of such a debriefing and the problem of consumer inattention seem to argue against this technique for widespread belief change.

Hypothesis-testing. Closely related to both the ideas of setting up a new illusory correlation and the process debriefing paradigm is that of presenting the subjects with some new hypothesis to test. Particularly if the experimenter can provide some hypothesis that incorporates an explanation for the subjects' previous beliefs with some new direction or framework regarding what they "should" believe, it seems that this technique should be effective in bringing about belief change. Given that subjects accept the new hypothesis for testing, they will then tend to collect hypothesis-confirming evidence. From this hypothesis-testing experience, subjects should attain some realization of why they believed as they did in the past and why they should change their beliefs in light of new evidence. In effect, the subjects will be doing a form of process debriefing and will also be building up a new illusory correlation on their own, with some structure provided by the experimenter.


Certainly the areas of belief persistence and belief change are fascinating areas for theoretical study. In addition, these phenomena have clear implications in everyday life, especially for marketers. It is hoped that this review of the findings related to the cognitive explanation for belief persistence will yield some insights and will generate future research in the development of cognitively-based approaches to belief change.


Anderson, Craig A., Lepper, Mark R., and Ross, Lee (1980), "Perseverance of Social Theories: The Role of Explanation in the Persistence of Discredited Information," Journal of Personality and Social Psychology, 39, No. 6, 1037-49.

Borgida, Eugene and Nisbett, Richard E. (1977), "The Differential Impact of Abstract vs. Concrete Information on Decisions," Journal of ApPlied Social Psychology, 7, 258-71.

Chapman, Loren J. (1967), "Illusory Correlation in Observational Report," Journal of Verbal Learning and Verbal Behavior, 6, 151-55.

Cohen, Claudia E. (1981), "Person Categories and Social Perception: Testing Some Boundaries of the Processing Effects of Prior Knowledge," Journal or Personality and Social Psychology, 40, 441-52.

Ehrlich, H. J. (1973), The Social Psychology of Prejudice, New York: John Wiley and Sons.

Hamilton, David L. and Gifford, Robert K. (1976). "Illusory Correlation in Interpersonal Perception: A Cognitive Basis or Stereotypic Judgments," Journal of Experimental Social Psychology, 1, 392-407.

Hamilton, David L. and Rose, Terrence L. (1980), "Illusory Correlation and the Maintenance of Stereotypic Beliefs," Journal of Personality and Social Psychology, 39, 832-45.

Moreland, Pamela (1982), "P&G Trying to Exorcise Devil Rumor," Los Angeles Times, January 77.

Ross, Lee, Lepper, Mark R., and Hubbard, Michael (197;), "Perseverance in Self-Perception and Social Perception: Biased Attributional Processes in the Debriefing Paradigm," Journal of Personality and Social Psychology, 32, 880-92.

Ross, L., Lepper, M. R., Strack, F., and Steinmetz, J. (1977), "Social Explanation and Social Expectation: Effects of Real and Hypothetical Explanations on Subjective Likelihood," Journal of Personality and Social Psychology, 35, 817-29.

Rothbart, Myron, Evans, Mark, and Fulero, Solomon (1979), "Recall for Confirming Events: Memory Processes and the Maintenance of Social Stereotypes," Journal of Experimental Social Psychology, 15, 343-55.

Smedslund, J. (1963), "The Concept of Correlation in Adults," Scandinavian Journal of Psychology, 4, 165-73.

Snyder, Mark (1981), "Seek, and Ye Shall Find: Testing Hypotheses About Other People," in Social Cognition: The Ontario Symposium on Personality and Social Psychology, eds. E. T. Higgins, C. P. Herman, and M. p. Zanna, Hillsdale, NJ: Erlbaum Associates.

Snyder, Mark and Uranowitz, Seymour W. (1978), "Reconstructing the Past: Some Cognitive Consequences of Person Perception," Journal of Personality and Social Psychology, 36, 941-50.

Tybout, Alice M., Calder, Bobby J., and Sternthal, Brian (1981), "Using Information Processing Theory to Design Marketing Strategies," Journal of Marketing Research, 18, 73-79.

Ward, W. C. and Jenkins, H. M. (1965), "The Display of Information and the Judgment of Contingency," Canadian Journal of Psychology, 19, 231-41.