On the Use of Naive Theories of Bias to Remove Or Avoid Bias: the Flexible Correction Model

ABSTRACT - In recent years, much attention in social psychology has focused on avoidance or removal of bias. The Flexible Correction Model (FCM; Wegener and Petty, 1997) conceptualizes such attempts as guided by individuals’ naive theories of bias (i.e., individuals’ beliefs and perceptions regarding the direction and magnitude of potential biases relevant to a given judgment). This paper introduces the core of the FCM, differentiates it from past models of bias correction, presents empirical support for theory-based correction, and addresses recent alternative explanations for FCM studies.

Citation:

Duane T. Wegener and Richard E. Petty (2001) ,"On the Use of Naive Theories of Bias to Remove Or Avoid Bias: the Flexible Correction Model", in NA - Advances in Consumer Research Volume 28, eds. Mary C. Gilly and Joan Meyers-Levy, Valdosta, GA : Association for Consumer Research, Pages: 378-383.

Advances in Consumer Research Volume 28, 2001     Pages 378-383

ON THE USE OF NAIVE THEORIES OF BIAS TO REMOVE OR AVOID BIAS: THE FLEXIBLE CORRECTION MODEL

Duane T. Wegener, Purdue University

Richard E. Petty, Ohio State University

ABSTRACT -

In recent years, much attention in social psychology has focused on avoidance or removal of bias. The Flexible Correction Model (FCM; Wegener and Petty, 1997) conceptualizes such attempts as guided by individuals’ naive theories of bias (i.e., individuals’ beliefs and perceptions regarding the direction and magnitude of potential biases relevant to a given judgment). This paper introduces the core of the FCM, differentiates it from past models of bias correction, presents empirical support for theory-based correction, and addresses recent alternative explanations for FCM studies.

Over the years, a number of potential "biases" have been studied in consumer settings. For example, attributes of products have been assimilated to the qualities of previous information encountered prior to the ad (e.g., Yi, 1990), evaluations of products have been influenced by the emotional qualities of programs in which ads occur (e.g., Gardner, 1985; Petty, Schumann, Richman, and Strathman, 1993), and evaluations have also been influenced by perceptions of the ad itself aside from perceptions of product attributes (Brown and Stayman, 1992). Many of these "biases" occur with little or no awareness on the part of consumers. What happens, however, if people ealize that one or more of these biasing factors might be influencing their perceptions of products? In many settings within and outside consumer behavior, people might go out of their way to avoid "bias." Sometimes this might occur because of external influences, as when people are explicitly warned not to use particular information (e.g., when judges instruct juries to disregard inadmissible evidence, see Fleming, Wegener, and Petty, 1999; Wegener, Kerr, Fleming, and Petty, in press). At other times, people might attempt to "correct" for bias without being asked to do so (e.g., because of a salient bias or personal concerns about bias, DeSteno, Petty, Wegener, and Rucker, 2000; Dunton and Fazio, 1997). In the current paper, we discuss the Flexible Correction Model (FCM; Wegener and Petty, 1997) as a general model of bias correction. The FCM relies on peoples’ use of their own perceptions of the bias at work in a given judgment setting, and this role for theories of bias makes the resultant corrections more flexible than in previous theories.

THEORETICAL BACKDROP

When we began work on the FCM in 1991, few models of bias correction existed. The most well-developed of these models were in the priming and categorization literatures. Taken together, these models might be called "partialling" models. The first of these, the set-reset approach, was introduced by Martin (1986). Consistent with the traditional priming stimuli, Martin (1986) began by assuming that representations of a target might include both positive and negative elements. When a concept or reaction is primed (e.g., by unscrambling sentences containing the concept), there is some overlap between reactions to the prime and reactions to the target (called "setting"). "Setting" results in an "assimilation" bias in judgment with perceptions of the target being more similar to the primes than when no primes have been encountered. However, if social perceivers identify some of their reactions as being due to the primes rather than the target (e.g., when the priming procedure was recent and blatant, Martin, 1986), they attempt to "partial out" the reactions to the primes (called "resetting"). "Resetting" reduces the number of prime-consistent reactions attributed to the target, which can reduce the assimilative bias or remove the assimilative bias. Also, if the social perceiver mistakenly identifies true reactions to the target (i.e., elements present in the initial target representation) as being reactions to the context (i.e., the primes), "resetting" can produce a contrast effectBwith reactions to the target being less like the context than if no context had been encountered. For a schematic representation of "setting" and "resetting" and more thorough discussion, see Wegener and Petty (1997; cf., Martin, Seta, and Crelia, 1990).

Schwarz and Bless (1992) developed a similar model referred to as the "inclusion-exclusion" model. Like Martin’s (1986) concept of "setting," "inclusion" of information in a target representation is expected to result in assimilation (with judgments of the target being more similar to the included information than when that information is not encountered). Also, like "resetting," "exclusion" (subtraction) of information from one’s representation of the target is expected to lessen the impact of information that could have been "included" in the representation. Exclusion can lessen assimilation or even create contrast, assuming that the excluded information is sufficiently extreme to shift the reaction to the target. In addition, excluded information can create this contrast due to the exclusion itself (as in "resetting") or because the excluded information is used to define the scale anchor or to set up a standard of comparison (see Schwarz and Bless, 1992, for additional discussion).

These "partialling" theories have both beenviewed as correction models by regarding the exclusion and resetting processes as correcting for the overlap between the representation of the target and the context experience or information. The models share a great deal, including the assumption that assimilation is the "default" bias, and that corrections lessen this assimilation and can even lead to contrast. Because both models equate partialling or subtraction with higher levels of cognitive effort, these models couple judgmental outcome (i.e., assimilation versus contrast) with levels of effortful processing. For example, Schwarz and Bless (1992) noted that "the emergence of contrast effects requires extra processing steps, and more effort, than the emergence of assimilation effects" (p. 240). That is, the set-reset and inclusion-exclusion models equate assimilation with "uncorrected" and lower effort default outcomes and contrast with more effortful "corrected" outcomes (see Wegener and Petty, 1997, for additional discussion).

A number of potential biases in consumer settings have been thought similar to this "overlap" view of bias, and efforts to avoid such bias have fit reasonably well with partialling models of bias correction. For example, Meyers-Levy and Sternthal (1993) reasoned that information about previous businesses occupying the same building as a current restaurant might influence perceptions of the current restaurant. Moreover, Meyers-Levy and Sternthal reasoned that information about a different class of business (i.e., a previous clothing store) would be viewed as information that should not be used in evaluating the current restaurant (i.e., that people might correct for effects of such information). Results were generally consistent with these notions and with a "partialling" view of the corrections. That is, when the previous business was a restaurant (and influences might be viewed as appropriate), assimilation to the valence of the previous restaurant was found. However, when the previous business was a clothing store (and influences might be viewed as inappropriate), contrast was found among research participants disposed toward putting high levels of effort into information processing. Assimilation was found for participants disposed toward cognitive simplification. More recently, Meyers-Levy and Tybout (1997) used blatant primes similar to Martin (1986)Bbrief positive or negative news storiesBand argued that cognitive resources of the type studied by Martin et al. (1990) have separate effects at encoding of product information and at the time of product judgment.

Despite many reasonable predictions and effects that have come out of this approach, it seemed to us that there are significant limitations to the partialling views of correction when thinking about developing a general bias correction theory. For example, assimilation might not be the only type of "uncorrected" effect, even when contextual stimuli are capable of activating relevant concepts. For example, Herr (1986) found that extreme exemplars associated with hostility (e.g., Adolf Hitler) made an ambiguous target person seem less hostile (a contrast effect; see also Herr, Sherman, and Fazio, 1983; Wegener, Petty, and Dunn, 1998). As noted by Petty and Wegener (1993), such effects need not rely on "corrective" exclusion or resetting but could be "uncorrected" biases and could be based on comparisons with an extreme anchor or standard (see Moskowitz and Skurnik, 1999; Stapel, Koomen, and Velthuijsen, 1998). If contrast is a possible default or "uncorrected" bias, then the partialling models are ill-equipped to serve as general correction theories (because they only deal with corrections for assimilative biasesB "setting" or "inclusion;" see Petty and Wegener, 1993; Wegener and Petty, 1997).

As a result, Wegener and Petty (1997; see also Petty and Wegener, 1993; Wegener and Petty, 1995; Wegener et al. 1998) developed the FCM as a more general model of bias correction. The FCM holds that corrections are aimed at removing (avoiding) the bias that social perceivers believe are associated with the factor(s) at hand. Consider, for example, a situation in whicha person realizes that his or her views of a political candidate might be unduly influenced by a particular political advertisement. The direction of the perceived bias should influence corrections that are made. That is, if the perceiver believes that the bias is to make the candidate seem negative because the ad categorizes him or her with a disliked politician (e.g., a politician rocked by recent scandal), the perceiver would likely adjust assessments of the target to be more positive. If the perceiver believes that the bias is to make the politician seem better than usual, because he or she is clearly not as bad as the scandalized politician and might be contrasted with him or her, the perceiver would adjust assessments of the target politician to be more negative. The perceived magnitude of the bias should also matter. If the perceiver believes there is a small bias, he or she would adjust assessments of the true qualities of the target less than if he or she believes that there is a large bias.

THE FLEXIBLE CORRECTION MODEL

The first four postulates of the FCM have guided much of the initial work on the model, and those postulates can be summarized as follows (see Wegener and Petty, 1997, for more detail and discussion).

1. Across situations and perceivers, there is variation in uncorrected effects.

2. Efforts at correction depend on motivation and ability to A) identify potential biases and B) to correct for those biases.

3. Identification of bias and bias correction itself are guided by naive theories of bias held and/or generated by the social perceiver.

A few comments serve to differentiate the FCM from the previous "partialling" models. Whereas the partialling models deal with "default" assimilative biases, the FCM assumes that people and situations differ in the extent to which the uncorrected effect is assimilative, nonexistent, or contrastive. For example, some people might interpret information about a target using an accessible construct, but others might spontaneously compare the target with a well-known exemplar available in the same setting. One situation might provide a universally salient exemplar against which targets are contrasted, whereas the same situation absent that salient exemplar might dispose most social perceivers toward interpreting target information in a context-consistent way. Whatever the source(s) of variation might be, the FCM acknowledges this variability and separates it from the processes of correction. That is, a given correction might operate similarly even if instigated across differing magnitudes or directions of "uncorrected" bias (see later examples of empirical studies). This can be compared with the partialling model assumptions of default assimilation and amount of correction (and the size of the default bias) being proportional to the amount of overlap between reactions to the context and to the target.

Many factors could influence motivation or ability to engage in corrections. Some motivations are quite global, such as motivations to engage in thoughtful activities (e.g., need for cognition, Cacioppo and Petty, 1982). Others might relate more specifically to a particular bias, such as motivation to avoid "prejudiced" responses (e.g., Dunton and Fazio, 1997), and many such motivations could be enhanced or diminished by situations themselves (e.g., presence or absence of a television program that discusses racial bias in legal settings). From the FCM point of view, people could also identify a bias but be unmotivated to correct for it because the bias is viewe as legitimate or even necessary (Petty, Wegener, and Fleming, 1996; see Wegener and Petty, 1995, 1997).

When people identify potential bias and are motivated and able to engage in corrections, they attempt to adjust target assessments in a direction opposite to the perceived bias and in a magnitude commensurate with the perceived size of the bias. Therefore, corrections can go in different directions or in different amounts, depending on perceptions of the bias at work. Of course, this corrective attempt does not take place in a vacuum. That is, available information (in memory or in the judgment setting) might often be consulted in the process of correcting. Part of the theory-guided correction is likely to include seeking of information (in memory or the environment) that might support these "corrected" assessments. If no such supporting information can be found, this might be one factor that could undermine theory-based correction. Also, some efforts at correction might involve information seeking and scrutiny to a greater extent than other corrections (see Wegener and Petty, 1997).

We assume that corrective processes often ensue when people become aware of a potential bias (and are motivated and able to engage in corrections). [This does not mean that perceivers must be able to consciously report correction processes they undertake, though people might often be able to report their perceptions of bias associated with a given context and target. Some studies that ask participants whether or not a context affected them face the ambiguity that a person might report a lack of a biasing effect because of corrections rather than because of a lack of perception of potential bias.] Therefore, corrections for bias need not occur only after reacting to the target. People might also anticipate a bias and attempt to avoid it (see Wegener and Petty, 1997; Wilson, Houston, and Meyers, 1998). Before people have a great deal of experience with a given type of bias, corrections would likely depend on some level of conscious awareness of the potential bias. However, with more experience of the bias and related corrections, less conscious awareness might be sufficient for instigating the corrections (and the correction itself might become routinized; Wegener and Petty, 1997). Even if awareness of the bias is quite conscious, the whole of the correction process would not generally be consciously reportable (consistent with Nisbett and Wilson, 1977). For example, people might often be unable to report which theory(ies) were used, even if the content of the relevant theory(ies) is "explicit" (see Wegener and Petty, 1998).

INITIAL EVIDENCE

One clear comparison between the FCM and the previous partialling models concerns the FCM’s assertion that people can make theory-based corrections for perceived contrastive bias (i.e., where the target is viewed as less similar to the context that when no context has been encountered). Obviously, a correction for contrast shifts judgments to be more like the context than when no correction occurs. Therefore, correction for contrast is in a direction opposite to the partialling of overlap between reactions to contextual stimuli and representations of targets. In a typical correction-for-contrast study, extreme exemplars are rated on some dimension, followed by ratings of more moderate targets either immediately or after an instruction not to let ratings of the contextual stimuli influence perceptions of the targets. Such instructions maximize the likelihood that correction will occur, but they do not specify whether people should be correcting in one direction or another. If corrections are guided by theories of bias (pretested prior to the study as being beliefs in contrast bias), then corrections should make assessments of targets more like assessments of the contextual stimuli. If, however, corrections consist of partialling for potential overlap between reactions to contexts and reactions to targets, then corrections should either leave target assessments unchanged (if no overlap is perceived) or should make target assessments less like the context (if overlap is perceived and removed).

Wegener et al. (1998) conducted one such study using stimuli similar to those used by Herr (1986). That is, Wegener et al. asked research participants to rate three violent people (e.g., Josef Stalin, Adolf Hitler, Saddam Hussein) or three non-violent people (e.g., Gandhi, the Poe, Jesus) and then to rate George Foreman and Arnold Schwarzenegger (either with or without a correction instruction). Both Foreman (an active boxer, but also an active product endorser at the time of the study) and Schwarzenegger (an actor in many violent movies, but also the spouse of Maria Shriver and a representative for the Presidential Physical Fitness program) were relatively ambiguous with respect to hostility. Therefore, it seemed that people might believe there is some overlap between the context and target people. However, pretests had shown that people believed the violent context (contemplating Stalin) would make the targets seem less violent than usual and that the nonviolent context (contemplating Jesus) would make the targets seem more violent than usual (i.e., perceived contrast).

Consistent with the Herr (1986) result noted earlier, participants rated the targets as less violent when they followed the violent rather than nonviolent context people (a contrast effect), so research participants happened to be directionally correct in their beliefs about bias. Presumably because they used these beliefs when asked to correct, however, the correction instruction led target ratings to become more like the context than when no correction instruction was given (e.g., targets were viewed as more violent when providing corrected rather than uncorrected ratings after contemplating HitlerCconsistent with adjustments for the perceived contrastive bias; see also Wegener and Petty, 1995; Stapel, Martin, and Schwarz, 1998). This correction was in a direction opposite to corrections that would have occurred if people correct by partialling out overlap between context and target.

In the Wegener et al. (1998) violence-rating experiment, opposite corrections (some more violent, some less violent) occurred when opposite theories (one suggesting too little perceived violence, the other suggesting too much perceived violence) were associated with two different contexts (see also Wegener et al., 1998, Study 1). Opposing corrections have also been associated with opposing theories about the impact of the same context on different targets (i.e., attractive models making average women appear less attractive, but making endorsed products appear more desirable, Wegener and Petty, 1995, Studies 2 and 3). [To the extent that theories of bias are generally beliefs in assimilative biases in the contexts where partialling models have been studied (e.g., see Petty and Wegener, 1993), the FCM is also capable of accounting for such corrections for perceived assimilation.] Different people can also hold opposing theories about the impact of the same context on the same targets (Wegener and Petty, 1995, Study 4). Moreover, people who perceive greater bias have been found to correct to a greater extent than people who perceive less bias (Wegener and Petty, 1995, Study 4), and corrections consistent with naive theories of bias have been found when only subtle instructions are used (Petty and Wegener, 1993) and when no correction instructions are present (DeSteno et al. 2000; Fleming et al. 1999).

In addition to evidence supporting theory-based correction, the importance of motivation and ability to identify bias has also gained support. For example, DeSteno et al. (2000) found people high in need for cognition assimilating judgments to experienced emotion when the bias was difficult to identify (e.g., because a separate-study paradigm was used and no manipulation checks on emotion occurred prior to target judgments), but found people high in need for cognition contrasting judgments to emotion when the bias was easy to identify (because an emotion manipulation check immediately preceded target judgments). Also, Stapel, Martin, and Schwarz (1998), used a more "conditional" correction instruction than in the original research (to correct only if a bias was detected) and found that participants were more likely to correct if the bias itself was blatant rather than subtle. Of course, this is entirely consistent with the FCM. [Stapel, Martin, and Schwarz (1998) seemed to agree with the FCM that likelihood of identifying a bias moderated the likelihood of theory-based corrections (p. 805), but believed that corrections following blatant instructions were based on the uncorrected judgments themselves rather than associated theories of bias (p. 803; the "valence-switching" alternative discussed in the next secion). This approach has difficulty accounting for existing FCM research (see next section), and theories of bias provide a more parsimonious explanation. Rather than referring to different processes depending on whether participants detect bias on their own or are directed to it, it seems quite possible that theory-based corrections took place in both cases.] The extent to which the biasing factor is blatant versus subtle should influence ability to detect the bias. Also, different correction instructions could influence motivation to search for bias and could affect the thresholds used in identifying a reaction as being "biased."

RECENTLY GENERATED ALTERNATIVES TO THEORY-BASED CORRECTION

Over the years, a number of potential alternative explanations of correction studies have been generated (either in print or in less formal discussions).

Discounting or recomputation.

One such alternative likens correction to processes in which people are hypothesized to set aside an unwanted response, ignore the biasing factor, and provide an independent response based on alternative inputs. For example, Schwarz and Clore (1983) proposed that people can discount their mood as an input to life satisfaction judgments if attention is brought to a cause of that mood (e.g., the weather or an unusual experimental chamber). [Interestingly, this discounting of mood was supposed to be most likely for negative rather than positive moods (but see Gorn, Goldberg, and Basu, 1993).] In such circumstances, people are supposed to compute satisfaction judgments based on inputs other than current feelings (e.g., satisfaction with work, with one’s spouse, etc.). Similar processes have been hypothesized for avoidance of category-based responses in impression formation, where people have been characterized as replacing the category-based reaction with piecemeal integration (attribute-by-attribute analysis) of individuating information about the target (e.g., Fiske and Neuberg, 1990). In most correction studies (and discounting or recomputation studies), the alternative inputs or individuating information would be the same across conditions because the information about the target is the same. Therefore, recomputation hypotheses would be for such processes to reduce or eliminate biases, not to reverse them. As we have seen in a number of correction studies, however, corrections often result in "overcorrections" wherein biases opposite to the suspected bias are created. For example, in one study (Petty, Wegener, and White, 1998), people who corrected for possible source effects on their judgments were more influenced by a dislikable than a likable source (see also DeSteno et al. 2000; Martin et al. 1990; Wegener and Petty, 1995; Wegener et al. 1998). Therefore, although discounting or recomputation might sometimes occur, this type of process does not seem to be a likely alternative explanation of many correction studies.

Valence-switching strategy.

Another alternative to theory-based corrections is that correction instructions (especially blatant instructions) make people provide a rating of a different valence from their initial reactions (Stapel, Martin, and Schwarz, 1998). Though not originally designed to address this explanation, a number of existing FCM studies are not very consistent with this alternative. In particular, a number of studies have found that respondents engage in different corrections consistent with differing theories of bias even when they give the same uncorrected ratings. For example, Wegener et al. (1998, Study 1) asked participants to rate the size of ambiguous animals after either a set of extremely large or moderately large animals (cf. Herr et al. 1983). Unexpectedly, there was no effect of context on people’s uncorrected ratings of targets. However, pretests had shown that people believed that the two contexts would have opposite uncorrected effects (that the extremely large context would make people view the ambiguous targets as relatively small, but the moderately large context would make people view the ambiguous targets as relatively large). If the valence-switching alternative is correct, correction instructions would lead to similar corrections for each context (because the uncorrected reactions were the same across conditions). In contrast, theory-based corrections should push ratings in opposite directions. The results showed opposite corrections, with target judgments significantly higher following the correction instruction when the context was extremely large rather than moderately large; see Wegener et al. 1998).

Theory-based responding.

Another existing study addresses not only the valence- switching alternative, but also confronts another possibility: theory-based responding. That is, one might wonder whether participants in the studies described earlier have initial reactions to targets and then correct those assessments. An alternative would be that theories themselves suggest an appropriate response and that initial reactions are cast aside (similar to a portion of the discounting alternative). Yet, if corrections occur for perceived rather than for actual bias (and we have seen corrections consistent with theories of bias even when no bias actually exists in uncorrected settings), it would seem possible that similar corrections could take place regardless of whether an initial bias exists or not. If people have these initial reactions, and these initial reactions constitute a starting point from which corrections occur, then the same corrections could result in different final judgments when initial reactions differ.

Petty, Wegener, and White (1998) conducted an FCM-based study that addresses this possibility. The study was initially predicated on the idea that the amount of effort given to a persuasion task can vary, regardless of whether or not corrections occur in that setting (see FCM Postulate 6, Wegener and Petty, 1997). That is, one way to conceptualize the persuasion domain is to partition the literature into situations in which bias is not salient (where traditional models such as the Elaboration Likelihood Model, Petty and Cacioppo, 1986, organize that work) and in which bias is salient (where models such as the FCM might be brought to bear; see Wegener and Petty, in press, for a discussion of this partitioning in the context of mood effects on judgment). This partitioning does not attempt to change current conceptualizations of persuasion (such as the ELM). Rather, it notes that issues of potential bias correction have largely been ignored in previous work on persuasion.

Half of the Petty et al. participants engaged in a replication of many past studies of attitude change. Participants encountered either a likeable or dislikeable communicator presenting a persuasive message under either low- or high-processing conditions (a combination of high relevance and low distraction in high-processing conditions and low relevance and high distraction in low-processing conditions). Results for these conditions replicated past findings, with greater impact of source likeability under low- rather than high- processing conditions (cf. Chaiken, 1980). When participants were asked not to let any nonmessage factors influence their perceptions of the proposed policy, however, people who encountered the message under low-processing conditions were no longer influenced by source likeability. However, the same corrections occurred for people who encountered the message under high-processing conditions (where no impact of source likeability had occurred in uncorrected settings). This created the opposite bias whereby people provided more favorable ratings of the policy when it was presented by the dislikeable rather than the likeable source. Such a result suggests that people "started" with initial reactions because the same size and direction of correction led to different final judgments for people who had different initial (uncorrected) reactions (based on high versus low processing effort). Therefore, this initial work suggests that people take their current reactions into account during the correction process, rather than simply providing responses that are directly called for by the theories themselves. This result also addresses the earlier valence-switching alternative because different corrections occurred for the different sources even though uncorrected reactions were the same across sources in high-processing conditions. Finally, the FCM approach and the Petty et al. (1998) results also suggest that social perceivers can exert a great deal of effort attempting to seek "correct" assessments of a target without necessarily also being concerned about avoiding bias. In other words, seeking "correctness" does not necessarily imply avoiding "bias."

Therefore, it seems hat observed effects of correction instructions have been quite consistent with the theory-based corrections hypothesized by the FCM but have been somewhat at odds with past theories (i.e., the partialling approaches) and with recent alternatives.

CONCLUSIONS AND FUTURE DIRECTIONS

By conceptualizing corrections as driven by naive theories of bias, the FCM goes beyond past partialling views that were developed within the longstanding priming/construct accessibility tradition. One important way in which the FCM goes beyond these perspectives is by uncoupling judgment outcomes (i.e., assimilation versus contrast) from categories of judgment process (i.e., "corrected" versus "uncorrected;" "effortful" versus "noneffortful"). According to the FCM, assimilation or contrast can be relatively effortful or effortless and can be either "corrected" or "uncorrected" outcomes (Petty and Wegener, 1993; Wegener and Petty, 1995, 1997). These and other flexibilities inherent in the model open up a wide variety of potentially interesting and useful research directions (see also Wegener and Petty, 1997; Wegener et al., 1998).

Although this paper focuses on outlining the model and the existing support for it, there are substantial portions of the model that have not yet been thoroughly addressed. That is, the model postulates that use of a theory of bias in correction depends on accessibility of the theory, applicability of the theory, and utility of the theory in serving the perceiver’s judgment goals. The model also specifies that there can be variation in cognitive effort involved in both uncorrected and corrected judgments. Finally, similar to some dual-process theories in social psychology, such as the Elaboration Likelihood Model (Petty & Cacioppo, 1986; Petty & Wegener, 1999), the FCM specifies that increases in effort and elaboration involved in correction increases the likelihood of that corrected perception persisting over time, resisting future attempts at change, and influencing future thoughts and behavior. That is, although correction is often an additional step beyond uncorrected assessments, the amount of effort in that correction can vary and so can the consequences of that correction (see Wegener and Petty, 1997; Wegener, Dunn, and Tokusato, in press, for additional discussion). [Though corrections might generally require more effort than lack of corrections, this does not mean that "corrected" assessments should necessarily persist over time or resist changes more than "uncorrected" assessments (see Wegener, Dunn, and Tokusato, in press; see Petty and Wegener, 1998, for similar comments regarding comparison of changed and unchanged attitudes; see also Priester, Wegener, Fabrigar, and Petty, 1999).] It is our hope that the FCM serves to spark interest in widely applicable models that accentuate commonalities in process across research domains. We believe that such models hold great promise for integrating work in social psychology, consumer behavior, and related domains, and we look forward to similar future developments and integration.

REFERENCES

Brown, Steven P. and Douglas M. Stayman (1992), "Antecedents and Consequences of Attitude Toward the Ad: A Meta-analysis,"Journal of Consumer Research, 19 (June), 34-51.

Cacioppo, John T. and Richard E. Petty (1982), "The need for cognition," Journal of Personality and Social Psychology, 42 (January), 116-131.

Chaiken, Shelly (1980), "Heuristic Versus Systematic Information Processing and the Use of Source Versus Message Cues in Persuasion," Journal of Personality and Social Psychology, 39 (November), 752-756.

DeSteno, David, Richard E. Petty, Duane T. Wegener, and Derek Rucker (2000), "Beyond Valence in the Perception of Likelihood: the Role of Emotion Specificity," Journal of Personality and Social Psychology, 78 (March), 397-416.

Dunton, Bridget C. and Russell H. Fazio(1997), "An Individual Difference Measure of Motivation to Control Prejudiced Reactions," Personality and Social Psychology Bulletin, 23 (March), 316-326.

Fiske, Susan T. and Steven L. Neuberg (1990), "A Continuum of Impression Formation, from Category-based to Individuating Processes: Influences of Information and Motivation on Attention and Interpretation," In Advances in experimental social psychology, Vol. 23, ed. Mark P. Zanna, New York: Academic Press, 1-74.

Fleming, Monique, Duane T. Wegener, and Richard E. Petty (1999), "Procedural and Legal Motivations to Correct for Perceived Judicial Biases," Journal of Experimental Social Psychology, 35 (March), 186-203.

Gardner, Meryl P. (1985), "Mood States and Consumer Behavior: A Critical Review," Journal of Consumer Research, 12 (December), 281-300.

Gorn, Gerald J., Marvin E. Goldberg, and Kunal Basu (1993), "Mood, Awareness, and Product Evaluation," Journal of Consumer Psychology, 2 (3), 237-256.

Herr, Paul M. (1986), "Consequences of Priming: Judgment and Behavior," Journal of Personality and Social Psychology, 51 (December), 1106-1115.

Herr, Paul M., Steven J. Sherman, and Russell H. Fazio (1983), "On the Consequences of Priming: Assimilation and Contrast Effects," Journal of Experimental Social Psychology, 19 (May), 323-340.

Martin, Leonard L. (1986), "Set/reset: Use and Disuse of Concepts in Impression Formation," Journal of Personality and Social Psychology, 51 (September), 493-504.

Martin, Leonard L., John J. Seta, and Richard A. Crelia (1990), "Assimilation and Contrast as a Function of People’s Willingness and Ability to Expend Effort in Forming an Impression," Journal of Personality and Social Psychology, 59 (July), 27-37.

Moskowitz, Gordon B., and Ian W. Skurnik (1999), "Contrast Effects as Determined by the Type of Prime: Trait Versus Exemplar Primes Initiate Processing Strategies That Differ in How Accessible Constructs Are Used," Journal of Personality and Social Psychology, 76 (June), 911-927.

Meyers-Levy, Joan and Brian Sternthal (1993), "A Two-Factor Explanation of Assimilation and Contrast Effects," Journal of Marketing Research, 30 (August), 359-368.

Meyers-Levy, Joan and Alice M. Tybout (1997), "Context Effects at Encoding and Judgment in Consumption Settings: The Role of Cognitive Resources,"Journal of Consumer Research, 24 (June), 1-14.

Nisbett, Richard E., and Timothy D. Wilson (1977), "Telling More than We Can Know: Verbal Reports on Mental Processes," Psychological Review, 84 (March), 231-259.

Petty, Richard E., and John T. Cacioppo (1986), Communication and Persuasion: Central and Peripheral Routes to Attitude Change, New York: Springer-Verlag.

Petty, Richard E., David Schumann, Steven Richman, and Alan Strathman (1993), "Positive Mood and Persuasion: Different Roles for Affect under High and Low Elaboration Conditions," Journal of Personality and Social Psychology, 64 (January), 5-20.

Petty, Richard E., and Duane T. Wegener (1993), "Flexible Correction Processes in Social Judgment: Correcting for Context-induced Contrast," Journal of Experimental Social Psychology, 29 (March), 137-165.

Petty, Richard E., and Duane T. Wegener (1998), "Attitude Change: Multiple Roles for Persuasion Variables," in the Handbook of Social Psychology (4^th ed.), Vol. 1, ed. Daniel Gilbert et al., New York: McGraw-Hill, 323-390.

Petty, Richard E., and Duane T. Wegener (1999), "The Elaboration Likelihood Model: Current Status and Controversies," in Dual Process Theories in Social Psychology, eds. Shelly Chaiken and Yaacov Trope, New York: Guilford Press, 41-72.

Petty, Richard E., Duane T. Wegener, and Monique A. Fleming (1996), "Flexible Correction Processes: Perceived Legitimacy of Bis and Bias Correction," Sturbridge, MA: Society for Experimental Social Psychology annual meeting (October).

Petty, Richard E., Duane T. Wegener, and Paul White (1998), "Flexible Correction Processes in Social Judgment: Implications for Persuasion," Social Cognition, 16 (Spring), 93- 113.

Priester, Joseph R., Richard E. Petty, Duane T. Wegener, and Leandre R. Fabrigar (1999), "Examining the Psychological Processes Underlying the Sleeper Effect: the Elaboration Likelihood Model Explanation," Media Psychology, 1 (1), 27-48.

Schwarz, Norbert and Herbert Bless (1992), "Constructing Reality and its Alternatives: an Inclusion/exclusion Model of Assimilation and Contrast Effects in Social Judgment," in The Construction of Social Judgments, eds. Leonard L. Martin and Abraham Tesser, Hillsdale, NJ: Erlbaum, 217-245.

Schwarz, Norbert and Gerald L. Clore (1983), "Mood, Misattribution, and Judgments of Well-being: Informative and Directive Functions of Affective States," Journal of Personality and Social Psychology, 45 (September), 513-523.

Stapel, Diederik A., Willem Koomen, and Aart S. Velthuijsen (1998), "Assimilation or Contrast?: Comparison Relevance, Distinctness, and the Impact of the Accessible Information on Consumer Judgments," Journal of Consumer Psychology, 7 (1), 1-24.

Stapel, Diederik A., Leonard L. Martin, and Norbert Schwarz (1998), "The Smell of Bias: What Instigates Correction Processes in Social Judgments?," Personality and Social Psychology Bulletin, 24 (August), 797-806.

Wegener, Duane T., Meghan Dunn, and Danny Tokusato (in press), "The Flexible Correction Model: Phenomenology and the Use of Naive Theories in Avoiding or Removing Bias," in Cognitive Social Psychology: On the Tenure and Future of Social Cognition, ed. Gordon B. Moskowitz, Mahwah, NJ: Erlbaum.

Wegener, Duane T., Norbert Kerr, Monique A. Fleming, and Richard E. Petty (in press), "Flexible Corrections of Juror Judgments: Implications for Jury Instructions," Psychology, Public Policy, and Law.

Wegener, Duane T., and Richard E. Petty (1995), "Flexible Correction Processes in Social Judgment: the Role of Naive Theories in Corrections for Perceived Bias," Journal of Personality and Social Psychology, 68 (January), 36-51.

Wegener, Duane T., and Richard E. Petty (1997), "The Flexible Correction Model: the Role of Naive Theories of Bias in Bias Correction," in Advances in Experimental Social Psychology, Vol. 29, ed. Mark P. Zanna, Mahwah, NJ: Erlbaum, 141-208.

Wegener, Duane T., and Richard E. Petty (1998), "The Naive Scientist Revisited: Naive Theories and Social Judgment," Social Cognition, 16 (Spring), 1-7.

Wegener, Duane T., and Richard E. Petty (in press), "Understanding Effects of Mood Through the Elaboration Likelihood and Flexible Correction Models," in Theories of Mood and Cognition: A User’s Handbook, eds. Leonard Martin and Gerald Clore, Mahwah, NJ: Erlbaum.

Wegener, Duane T., Richard E. Petty, and Meghan Dunn (1998), "The Metacognition of Bias Correction: Naive Theories of Bias and the Flexible Correction Model," in Metacognition: Cognitive and Social Dimensions, ed. Vincent Yzerbyt et al., London: Sage, 202-227.

Wilson, Timothy D., Christopher E. Houston, and Jonathan M. Meyers (1998), "Choose Your Poison: Effects of Lay Beliefs about Mental Processes on Attitude Change," Social Cognition, 16 (Spring), 114-132.

Yi, Youjae (1990), "The Effects of Contextual Priming in Print Advertisements," Journal of Consumer Research, 17 (September), 215-222.

----------------------------------------