Be yourself, image is nothing: bias correction when viewing ads in sequence

Abstract

Attempts at social influence are frequently encountered in sequence rather than in isolation. We suggest that messages can play a new correction-based role when they prompt message recipients to consider possible biases. The resulting corrections can reduce or even reverse the typical context effects. Two experiments examined advertisements as contexts for activating corrections. In Experiment 1, the context ad contained a tagline highlighting potential endorser bias. In Experiment 2, a sequence of context ads featured the same endorser endorsing different products, subtly prompting the potential for endorser bias. In both studies, research participants corrected for the assimilative effect of the celebrity endorser more when motivation was relatively high rather than low. Implications for the practice of social influence are discussed.

Keywords::

• advertising
• assimilation
• correction
• motivation

Many everyday attempts at social influence are encountered within sequences of influence attempts. Yet, much research addresses isolated persuasive messages. Examining persuasive variables in an isolated message is often useful. However, considerable influences on message effectiveness might reside in the context of surrounding topics and influence attempts, and many of these effects have not been thoroughly addressed in the social influence literature. For example, consider the setting in which consumers encounter advertising. Such ads are often encountered as a series of radio, magazine, or TV commercials. In such settings, perceptions of the target product might not only be affected by persuasive variables in the target ad itself, but might also be biased by accessible concepts or ideas activated by the contextual ads.

Some studies have focused on how the ads preceding the target ad can make the perceptions of the target product assimilate toward (e.g., Yi, 1990) or contrast away from (e.g., Keller, 1991; Tormala & Petty, 2007) the stimuli exposed in the contextual ads. Assimilation to a context is more likely when consumers think about the potential similarity of context and target (Mussweiler, 2003), and when there is a feature match (Herr, 1986, 1989) or an overlap in the interpretive ranges (Chien, Wegener, Hsiao, & Petty, 2010) of the context and target. On the other hand, contrast of target perceptions away from the context can occur when the context is used as a comparison standard for the target (Herr, 1989), when consumers test the hypothesis that the context and target are dissimilar (Mussweiler, 2003), and when there is a mismatch in features between the context and target (Herr, 1989) or lack of overlap in the interpretive ranges of the context and target (Chien et al., 2010).

A new correction role for context messages

We believe that ads preceding a target ad can do more than serve as a context to create a default assimilation or contrast effect. Rather, we propose that contextual ads can also activate attempts to guard against possible biases in perceptions of the target ad. Of note, we suggest that corrective attempts to “recalibrate” product impressions can be triggered by stimuli within the ad setting. That is, consumers may engage in judgmental correction without external instructions attributed to sources outside the ads themselves (see also Handley et al., 2009).

For consumers to engage in effortful correction of product impressions in a marketing setting, the context ad(s) would often have to contain some feature(s) to facilitate consumers' recognition that an undue persuasive bias might have contaminated their target judgments. It might seem odd that marketers would devote advertising content to address effects of other ads rather than simply presenting the qualities of their own product. However, sometimes it could be quite important to combat the success of a competitor's ads by prompting corrections for the persuasive effects of the competitor's message. Unfortunately, in many research and applied settings, these potential roles for ads are ignored.

Despite the lack of research attention, some high-profile examples exist of ads that we would identify as likely to instigate corrections. Consider the “Image is Nothing” ads from Sprite (a soft drink brand) as an example. Similar to other beverage ads, the “Image is Nothing” ads typically used a celebrity endorser who obviously drank Sprite. However, the Sprite ad concluded with a tagline asserting that “Image is Nothing. Thirst is Everything. Obey Your Thirst!” This ad suggests to consumers that reactions to a celebrity are not relevant to the drink qualities and that they should choose Sprite not because of the celebrity but, instead, because it does what a drink is really supposed to do-quench thirst. We believe that a number of specific features of ad sequences might lead consumers to discount celebrity influences in subsequent ads.

If a consumer were to view the Sprite ad before seeing another ad that features a famous or attractive endorser, we suggest that the Sprite ad might undermine the influence of that favorable endorser by engaging corrections for “undesirable” influences of “image.” Under the right conditions, corrections might not only remove positive persuasive effects of endorser attractiveness; they might even make the attractive endorser backfire and lead to more negative (or at least less positive) reactions than if that persuasive variable had not been used (Kang & Herr, 2006; Petty, Wegener, & White, 1998).

For judgmental correction to occur, message recipients should not only perceive potential biases, they should also have sufficient ability and motivation to correct for the perceived biases (Wegener & Petty, 1995, 1997, 2001; Wilson & Brekke, 1994). This is because correction attempts, like other forms of meta-cognition, typically require a relatively high degree of thinking (see Petty, Briñol, Tormala, & Wegener, 2007). The current research focuses on the potential for contextual communications (advertisements) to alert people to the potential for bias and the role of motivation to think carefully about the target ad in generating corrections as a way to deal with that potential for bias. Previous work outside the social influence context has examined responses to blatant biasing factors and found motivation to think about the target as a key variable leading to corrections (e.g., Isbell & Wyer, 1999; Martin, Seta, & Crelia, 1990). However, this work has tended not to compare different levels of bias awareness. In work testing the Flexible Correction Model (FCM; Petty & Wegener, 1993; Wegener & Petty, 1995, 1997), experimenter instructions were used to make people aware of potential biases. However, these instructions likely also affected motivation to think about the target and to engage in corrections because they specifically asked research participants to avoid the identified biases. Therefore, we believed that it is important to demonstrate separable roles for variables that might prompt an awareness of the potential for bias and variables that would affect the amount of processing of information about the judgment target. Also, consistent with the FCM assumption that people possess or generate perceptions of the bias(es) at work, and that these perceptions guide correction efforts when motivated and able to think carefully, we collected pretest data ensuring that research participants believed that the relevant assimilative biases would operate in the settings under study.

Experiment 1

We began by investigating whether a tagline in a contextual ad could trigger judgmental correction in a subsequently encountered ad. In addition to the presence/absence of the tagline, we also manipulated motivation to think carefully about the target qualities. Those who received target ads without bias-relevant taglines should be vulnerable to the endorser-related bias, especially when motivation to process is low. However, when a contextual ad helped to identify endorsers as potentially biasing, corrections for endorser effects should be more likely with higher rather than lower motivation to think about target ads.

In order to directly relate the corrections to participants' theories of bias, we conducted a pretest in which theories of bias were measured. Nineteen pretest participants reported their perceptions of the judgmental effects of exposure to an attractive celebrity in an ad for a new chocolate. Responses were made on a nine-point scale from make the novel chocolate brand seem less favorable( − 4) to more favorable (4) than if no attractive endorser was present. Participants believed the celebrity endorser in the ad would make the chocolate brand seem more favorable (M = 2.05, SD = 1.27, t(18) = 7.06, p <  .001, in relation to 0).

Design and procedure

A total of 126 undergraduates participated in the experiment for extra credit in introductory marketing courses. Participants were run in two classroom sessions with roughly equal size, and were randomly assigned to a 3 (Tagline: tagline/no tagline/no tagline-no endorser control) ×  2 (Motivation: higher/lower) between-participants factorial design. We measured the endorser effects and the correction magnitudes by comparing higher- and lower-motivation control groups with experimental groups.

Participants in experimental groups received one of four experimental booklets. The first page informed participants that they would be reading ads excerpted from various magazines and their general attitudes toward products in the ads would be investigated. Participants were asked to read the ad as they would normally read ads in magazines or newspapers. Then, they received an ad for a fictitious shoe brand. Half of the participants were exposed to a tagline suggesting that endorser effects were biases to be avoided and the rest viewed the same ad without the tagline. Similar to Sprite's “Image is Nothing” slogan, the tagline in the contextual ad served to enhance participants' ability to identify potential endorser-based bias. In the tagline conditions, the ad-slogan included the statement “Other than you, whoever wears it is not important. Be yourself. Refuse to be affected by the endorser!” The tagline was intended to suggest that the person's own needs and ideals were the most important and that product endorsers were unduly biasing. This notion of undue influence should still be salient and accessible when the subsequent product was evaluated. In conditions without tagline, the rest of the layout of the shoe ad remained the same across all conditions. Participants' attitudes toward the fictitious shoes were assessed on the following page.

Participants then received a company-introduction passage similar to a brief magazine feature designed to manipulate process motivation in the subsequent target ad. In the higher-motivation condition, participants were told that a fictitious European chocolate brand was going to be introduced in the local area during the next month and that college students would be its target consumers. To further enhance motivation, recipients were told that their opinions and attitudes toward the target product were extremely important because only a small sample of students was providing their opinions and the data would be used to make critical decisions pertaining to the launch of the new product. On the other hand, when motivation was lower, participants were not given any statement about a product launch or responsibility for evaluating the product (cf. Petty, Cacioppo, & Schumann, 1983; Petty, Harkins, & Williams, 1980).

Finally, participants received the target ad for the fictitious chocolate brand endorsed by a favorite celebrity, M = 5.30, SD = 1.12 on a 1 = less favorable to 7 = more favorable scale in a pretest, significantly higher than the scale midpoint, t(29) = 6.37, p <  .001. The ad included a picture of chocolate and a short passage claiming a variety of flavors, such as strawberry, cherry, and milk. Participants in the control groups completed one of two booklets (i.e., higher or lower motivation), neither of which included an endorser for the target product. After the target ad, participants were asked to evaluate the target product and to answer the manipulation checks.

The target evaluation was the average of 3 seven-point scales (Cronbach's α = .84): bad/good, negative/positive, and unfavorable/favorable. Motivation when reading the target ad was the average of 3 seven-point semantic differential scales (Cronbach's α = .76): not/very important to answer accurately, not/very prudent to make the accurate answer, and not/very concentrating on providing the accurate answer. Participants' favorability toward the endorser was the average of 4 seven-point semantic differential scales (Cronbach's α = .90): unattractive/attractive, dislikeable/likable, negative/positive, and unfavorable/favorable. Participants' perceptions of the target information was the average of 2 seven-point scales (Cronbach's α = .82): not/very persuasive and with very weak/strong arguments. After an open-ended suspicion probe, participants were thanked, debriefed, and dismissed. No participants guessed the study purpose in the suspicion probe.

Results

Manipulation checks

A 3 (Tagline) ×  2 (Motivation) analysis of variance (ANOVA) on the motivation manipulation check yielded only the main effect of Motivation. In higher-motivation conditions, participants expressed greater motivation (M = 4.48, SD = 1.15) to make accurate judgments than participants in lower-motivation conditions (M = 3.63, SD = .94), F(1, 120) = 22.52, p (.001), η² = .16.

Participants' perceptions of information supporting the target were not strongly influenced by the manipulation of motivation, F(1, 120) = 2.85, p = .094, η² = .02. The perceived quality of the target information was non-significantly higher when participants were in higher-motivation (M = 4.01, SD = .94) rather than lower-motivation conditions (M = 3.68, SD = 1.08). More importantly, the quality of the target information was not perceived differently across the groups that did (M = 3.82, SD = 1.01) or did not (M = 3.86, SD = 1.03) receive the tagline in the context ad, F(1, 122) = .06, p = .80, η² <  .01.

Checks on endorser perceptions were based only on those conditions that had an endorser of the target product. Liking of the endorser was uninfluenced by the tagline (M = 5.10[SD = .98] vs. M = 5.02[SD = 1.13]), F(1, 88) = .12, p = .73, η² <  .01, or motivation (M = 5 .04[SD = .98] vs. M = 5.08[SD = 1.13]), F(1, 88) = .02, p = .88, η² <  .01, and there was no interaction, F(1, 88) = 1.77, p = .19, η² = .02. Therefore, any effects of the tagline cannot be attributed to changes in the evaluation of the endorser across conditions.

Target attitudes

A 3 (Tagline) ×  2 (Motivation) ANOVA on attitudes toward the target product revealed a Tagline main effect, F(2, 120) = 4.52, p = .01, η² = .07, and a Tagline ×  Motivation interaction, F(2, 120) = 4.28, p = .02, η² = .07. Of particular interest, an interaction contrast (Abelson & Prentice, 1997) focusing on the experimental conditions showed a significant Tagline ×  Motivation interaction, F(1, 88) = 4.89, p = .03, η² = .05. In the higher-motivation condition, assessments of the target were significantly lower following the tagline (M = 4.12, SD = .74), than when no tagline was encountered (M = 5.03, SD = .59), F(1, 88) = 13.31, p <  .001, η² = .13, see Figure 1. When motivation was lower, however, target attitudes did not differ between those who received a tagline (M = 4.80, SD = .95) and those who did not (M = 4.93, SD = 1.04), F(1, 88) = .27, p = .60, η² <  .01.

Figure 1 Target attitudes as a function of tagline and motivation in Experiment 1.

When comparing no-tagline conditions to control conditions, assimilation to the endorser was the general pattern. Collapsing across levels of motivation, participants' target attitudes were more favorable when receiving the endorser without a previous bias-relevant tagline (M = 4.98, SD = .84) than for those in the control group (M = 4.53, SD = 1.00), F(1, 76) = 4.76, p = .03, η² = .06. Consistent with past research on processing motivation, the endorser effects were significant when motivation was lower (celebrity endorser: M = 4.93, SD = 1.04; no endorser: M = 4.31, SD = 1.03, F(1, 76) = 4.45, p = .04, η² = .06) but not when motivation was higher (celebrity endorser: M = 5.03, SD = .59; no endorser: M = 4.75, SD = .95, F(1, 76) = .95, p = .33, η² = .01). However, though the interaction testing the differences in these effects was not significant, F(1, 76) = .64, p = .43, η² = .01.

After correcting for the perceived positive source bias in higher-motivation conditions, participants actually rated the target significantly less favorably (M = 4.12, SD = .74) than in the higher-motivation control group (M = 4.75, SD = .95), F(1, 76) = 4.65, p = .03, η² = .06. Thus, in higher-motivation conditions, participants appeared to have corrected for perceived assimilation when no significant bias had occurred, and this created a significant correction-induced contrast effect. That is, corrections led the celebrity source to backfire on the product. In lower-motivation conditions, when no significant corrections occurred, ratings following contextual taglines remained in the direction of assimilation (M = 4.80, SD = .95) compared with the lower-motivation control group (M = 4.31, SD = 1.03), though not significantly so, F(1, 76) = 2.75, p = .10, η² = .04.

Discussion

The results supported our proposal that the contextual ad with a bias-relevant tagline could significantly influence reactions to the target ad, especially in higher-motivation conditions. In fact, comparing with the no-endorser control groups, corrections even created a bias opposite to the typical (lower motivation) endorser effect. Our Motivation ×  Tagline design differed from past correction-related studies in two important ways. First, the correction prompts were included in the contextual ads. In past studies of correction for liked or attractive sources (e.g., Petty et al., 1998), research participants were reminded to consider bias-relevant knowledge in an activity external to the advertising setting. Including correction prompts in the contextual ads emphasizes the potential practical benefits of considering the use of ads to instigate corrections. Another key difference is that we made sure that the corrections occurred without influencing the perceived endorser favorability. This is important because past corrections for source attractiveness (e.g., Kang & Herr, 2006) are vulnerable to an alternative explanation that does not include correction at all. If the bias-identification activity decreased liking for attractive endorsers, then less favorable target judgments could reflect assimilation to the new, less positive views of the attractive endorser. In the current case, we ensured that research participants perceived the celebrity endorser equally positively across all conditions. Then, we found that significant negative shifts occurred only when the contextual ad contained a bias-relevant tagline in the higher-motivation condition. The current experiment also showed that corrections can be prompted by an ad that comes from an entirely different product category from the target. Therefore, marketers would have to monitor not only the competitors' ads from the same product category, but also other ads likely to appear in close proximity to the target ad, even if not in the same product category.

Experiment 2

The tagline in Experiment 1 was quite explicit (similar to the Obey Your Thirst ads). Even so, it only had a significant influence when motivation to think about the target was relatively high rather than low. It seems likely, however, that more subtle means of alerting people to the potential for bias might have conceptually similar effects. For example, a celebrity endorser may be more likely identified as a bias agent if she or he has been encountered frequently. With frequent endorsers, consumers may be more likely to suspect the motive of the endorser and attribute it to the money, instead of true support for the products. We let research participants view a sequence of ads in which the endorser frequently (vs. infrequently) endorsed different products prior to evaluations of the target product endorsed by the same celebrity. We assumed that the frequent endorsements would heighten participants' suspicion about the endorser and make participants more readily identify the same endorser appearing in the target ad as a potential source of bias. We also manipulated higher versus lower motivation to think carefully about the target in order to examine the effects of two limiting factors that should work in concert for triggering bias corrections (i.e., identify the biasing factor and possess higher motivation to carefully think about the target). We expected that bias corrections would be elicited when participants encountered the frequent endorser and when they had higher motivation to carefully think about the target.

The theory measurement pretest was similar to Experiment 1. Nineteen pretest participants reported perceptions of the proposed judgment effects of exposure to an attractive celebrity for reactions to a target notebook computer. Responses were made on a nine-point scale ranging from make the novel notebook brand seem less favorable ( − 4) to more favorable (4) than if no attractive endorser was present. Participants believed that the celebrity endorser would make the notebook brand seem more favorable than if the celebrity were not present (M = .97, SD = .91, t(18) = 4.53, p <  .001, in relation to 0).

Design and procedure

A total of 177 undergraduates (after dropping three incomplete responses) participated in sessions of 4–12 students. They were randomly assigned to a 3 (Celebrity Exposure Frequency [CEF]: high/low/none-control) ×  2 (Motivation: higher/lower) between-participant factorial design. Students received partial course credit in their introductory marketing classes. Target evaluations in the control groups served as the baseline against which the celebrity context effect (in lower-motivation conditions) and the judgmental correction (in higher-motivation conditions) were compared.

We used the same cover page and motivation manipulation as in our first study. Instead of evaluating each product immediately following each ad, participants rated the products after viewing all five ads in sequence. The first ad (for a fictitious notebook brand) served as the target ad, followed by four additional ads from different product categories, i.e., fruit juice, ring (jewelry), cellular phone, and coffee machine. The target ad presented a picture of a notebook topped with the brand name and a headline: “A mobile computer comes with piano mirror finish on the cover.” A favorable celebrity was either prominently placed on the target ad in the experimental conditions or not present in the control conditions. Participants in the experimental conditions were randomly assigned into either high or low level of CEF. In high-CEF conditions, the same favorable celebrity was repeatedly encountered (with different poses and outfits) in all but the third ad. However, in the low-CEF conditions, the same endorser was absent in all but the target ad. The celebrity used in the study generally endorses products related to personal appearance, such as body care, on print ads and TV commercials. In control conditions, no endorser was present in any of the ads.

After viewing all five ads, participants were asked to rate the target (along with another product in the third ad in order to minimize potential suspicion), complete manipulation checks, and respond to the suspicion probe. Nine-point scales (i.e., 1–9) rather than seven-point scales were used in all measures. An additional item, dislike/like, was included in the attitude scales along with items used in our first study (Cronbach's α = .91). The motivation check items were the same as in our first study (Cronbach's α = .89). Participants' positivity toward the celebrity endorser was measured by two scales: like/dislike and positive/negative (r = .89, p <  .01). Participants were also asked to recall the number of times the given celebrity was exposed in five ads. Finally, the perceived suitability of the celebrity to the target product was measured by three scales: fit/unfit, suitable/unsuitable, and relevant/irrelevant (Cronbach's α = .93).

Results

All dependent measures in the subsequent analyses were subject to 3 (CEF) ×  2 (Motivation) ANOVAs unless otherwise specified.

Manipulation checks

The ANOVA on the motivation check exhibited only a Motivation main effect, F(1, 171) = 47.53, p <  .001, η² = .22. More highly motivated participants reported that it was more important for them to make accurate target judgments (M = 5.94, SD = 1.66) than their lower-motivation counterparts (M = 4.23, SD = 1.65).

The celebrity was perceived more favorably than the neutral midpoint (M = 5.87, SD = 1.74, t(118) = 5.44, p <  .001). No other significant effects were found (Fs <  1). In particular, mean positivity toward the celebrity did not differ between highly motivated participants who are high (M = 5.73, SD = 1.98) versus low in CEF (M = 6.03, SD = 1.42), F(1, 115) = .43, p = .51, η² <  .01. Such parity in the ratings of celebrity favorability suggests that shifts in judgments associated with CEF are not attributable to shifts in favorability of the endorser.

A 2 (CEF: high/low) ×  2 (Motivation) ANOVA showed that participants encountering the celebrity four times recalled more endorser exposures (M = 4.12, SD = .49) than participants only seeing the celebrity in the target ad (M = 1.03, SD = .18), F(1, 115) = 2019.18, p <  .001, η² = .95. No other effects were found (Fs <  1).

Finally, the celebrity was perceived as significantly less well-suited to the target than the scale midpoint (M = 3.88, SD = 1.64), t(118) = − 7.49, p <  .001, consistent with the notion that the source was an illegitimate piece of information and thus unlikely to serve as an argument to support the quality of the product (cf. Kang & Herr, 2006; Petty & Cacioppo, 1984). No significant differences occurred across the experimental conditions (Fs <  1.25). Of note, higher motivation/high-CEF participants perceived the celebrity as equally unfit/unsuitable to the target (M = 3.66, SD = 1.92) as their higher motivation/low-CEF counterparts (M = 3.76; SD = 1.04, F(1, 115) = .06, p = .81, η² <  .01. Accordingly, differences between high- and low-CEF conditions were not due to differences in the perceived suitability of the celebrity to the product, just as they were not due to shifts in the perceived favorability toward the celebrity.

Target attitudes

A significant main effect of CEF was observed, F(2, 171) = 8.55, p <  .001, η² = .09.Overall, low-CEF participants were more favorable toward the target product (M = 5.75, SD = 1.11) than participants in control (no celebrity) conditions (M = 4.70, SD = 1.58, F(1, 174) = 15.81, p <  .001) and in high-CEF conditions (M = 5.06, SD = 1.51, F(1, 174) = 7.13, p <  .01). The Motivation main effect did not approach significance, F(1, 171) = 1.98, p = .16, η² = .01. More importantly, the CEF ×  Motivation interaction was significant, F(2, 171) = 3.53, p = .03, η² = .04. CEF negatively influenced target evaluations (high-CEF, M = 4.58, SD = 1.13; low-CEF, M = 5.58, SD = .86, F(1, 171) = 7.56, p <  .01, η² = .04) in the higher-motivation condition, but not when motivation was lower (high-CEF, M = 5.53, SD = 1.70; low-CEF, M = 5.92, SD = 1.30), F(1, 171) = 1.14, p = .29, η² <  .01 (see Figure 2).

Figure 2 Target attitudes as a function of celebrity exposure frequency and motivation in Experiment 2.

The interaction between CEF and motivation sustains our suggestion that correction is more likely when accuracy motivation is higher, and the feature of ad sequence (i.e., high-CEF in a relatively short exposure interval) triggers bias awareness. Similar to Experiment 1 and consistent with much previous FCM research, correction was apparently driven by perceptions of bias rather than actual bias. That is, compared with the low-motivation control condition (M = 4.50, SD = .89), the celebrity positively influenced target evaluations both when CEF was low (M = 5.92, SD = 1.30), F(1, 171) = 15.07, p <  .001, η² = .08, and when CEF was high (M = 5.53, SD = 1.70), F(1, 171) = 8.02, p <  .01, η² = .05.However, when motivation was higher, high-CEF led to lower evaluations than low-CEF (reported earlier) even though the low-CEF conditions no longer differed from the higher-motivation control condition (p > .19; cf. Petty et al., 1998).

Discussion

Consistent with Experiment 1, corrections were most likely when ad recipients were motivated to think carefully about the target ad and become aware of the potential bias (e.g., encountering the same and ill-suited endorser frequently). These corrections were also consistent with the naïve theories of bias that pretest participants reported. It is important to note that the differences in target evaluations were not simply because the frequent endorsement of products undermined ad recipients' favorability toward the endorser. Instead, they continued to view the endorser favorably, but corrected their target perceptions to reduce the impact of the endorser on product evaluations.

General discussion

The current research suggests that ads can play a new correction role to influence reactions by ad recipients. As a part of the advertising context, an ad can make consumers aware of potential biases that influence perceptions of a target brand. This can occur even when the potential for bias is related to an ad for a different type of product or to ads for a whole variety of unrelated products. When this awareness is accompanied by sufficient motivation to think about the target, the suspected biases can be reduced, eliminated, or even reversed compared with circumstances in which such awareness or motivation is lacking. Therefore, if one's competitor frequently uses an advertising strategy that can be cast as “biasing” to consumers (e.g., the use of celebrity), alerting consumers to the biasing nature of that factor (celebrity) can actually lead that advertising feature to backfire on the competitor.

Attempts to correct for perceived biases sometimes make target judgments more “accurate” and sometimes less accurate than when no correction occurs (as in higher-motivation conditions of Experiment 1). The problem with perceptions of bias is that such perceptions are often poorly calibrated. Perceived biases can be larger or smaller than actual biases (see Wegener & Petty, 1995, 2001). If perceived biases are larger than the actual biases (including when none actually exists), corrections can create a bias opposite in direction to the perceived bias. If perceived bias and actual bias match in direction and are close in magnitude, however, then corrections can make judgments more accurate.

The current results are consistent with theory-based correction leading to the observed outcomes, but future research could further clarify the mechanisms at work. For example, the motivation and awareness variables interacted in both experiments. The results fit with the notion of motivation being necessary to engage corrections when aware of the potential for bias. This awareness of the potential for bias could be present across levels of motivation (when a tagline was previously encountered or the endorser was exposed frequently). Alternatively, the higher level of motivation might have enhanced the extent to which the alert to the possibility of bias was effective in getting people to detect the potential for bias in the first place (perhaps in addition to being necessary to undertake the corrections). Without an independent check on bias awareness (bias detection) per se, we cannot tell which of these possibilities is true in the current data. However, either of these possibilities is consistent with the notion that both bias awareness and motivation to think come together to engage attempts at bias correction.

Another area where future work could help to clarify the mechanisms underlying the current results concerns perceptions of the sources in both experiments. We measured the overall favorability of participants toward the source (in Experiment 1) and the suitability of the endorser to the product (in Experiment 2). The consistent evaluations of the source suggest that the results were more likely due to corrections of the perceived positive influences of the sources rather than a change in evaluation of the source across conditions. However, some previous research has shown that frequency of endorsement can create more negative views of an endorser's attractiveness and credibility (Tripp, Jensen, & Carlson, 1994). Because credibility would seem to be correlated with overall evaluations (in both experiments) and, perhaps especially, with judged suitability (in Experiment 2), we think it is unlikely that there were shifts in perceived credibility that we missed with our measures. We did not directly measure credibility in our experiments, however, so shifts in perceived credibility (unrelated to overall source evaluation or suitability) might remain a possibility. The effects of endorser exposure frequency could open up a variety of interesting questions, however, including additional moderators of the exposure frequency effects. More generally, many features of the ad sequence could determine the extent to which particular biases become salient.

The current research also has potential to provide new insights for practitioners of social influence. For example, when a marketer knows that a competing brand uses a particular advertising strategy such as celebrity endorsement, one might undermine the effectiveness of the competitor's ads by convincing consumers that influences of the celebrity are “biasing” by incorporating a bias-prompt tagline in one's own ad.

The current study on endorser frequency provides another example for the possibility that the context ads might invoke bias corrections. When studied in an isolated ad, a well-liked, attractive endorser would often have positive effects on product perceptions, though this effect might be reduced for some products when the processing motivation of that ad is higher rather than lower (Petty et al., 1983). In the current study, however, reduced endorser effects were more marked when the endorser had also been encountered in many ads within the sequence. On some level, this effect would seem to conflict with conventional wisdom and practice. That is, when the same endorser is ever-present in the media, the person would seem to be more popular and perhaps even more likely to bring positive responses to one's product. However, there may be situations in which this typical positive effect does not hold true. In particular, when the popularity of a target-irrelevant endorser results in many endorsements in a short period (or perhaps spaced across hours or days), highly motivated customers may correct for the otherwise positive effects of the endorser. These questions remain for future research. We hope that the current research highlights the potential importance of bias correction in persuasion settings, and we look forward to continued growth in the level of attention given to such matters.

Acknowledgements

We thank Duane Wegener and Richard Petty for comments on previous drafts of this manuscript.