Self‐concept clarity and responsiveness to false feedback

Abstract

We examined the extent to which people high and low in helpful and honest self‐concept clarity use labeling to guide decisions about their behavior. High helpful self‐concept clarity participants in Experiment 1 who received feedback indicating they were helpful were more likely to assist a confederate who dropped some pamphlets than were those not receiving this feedback. Low helpful self‐concept clarity participants showed no change in behavior as a result of the labeling. A similar pattern was found in Experiment 2 when participants received feedback indicating they were high in honesty and then presented with an opportunity to admit they received too many credits for their participation. In a partial replication of the first experiment, high helpful self‐concept clarity participants in Experiment 3 rated themselves as more helpful after receiving the feedback, but low helpful self‐concept clarity participants did not. The pattern across all three experiments suggests people with a clear, well‐articulated self‐concept are more likely to rely on self information to guide their behavior than those low on this personality dimension. The implication for social influence researchers is that this is an individual difference characteristic that moderates susceptibility to labeling techniques.

Research on the self‐concept demonstrates that individuals vary not only in terms of what they think of themselves, but also in the clarity and certainty of their self‐images (Baumgardner, 1990; Campbell, 1990; Campbell & Lavallee, 1993; Campbell et al., 1996; Setterlund & Niedenthal, 1993). Formally called self‐concept clarity, this individual difference is concerned with the extent to which one's self‐concept is “clearly and confidently defined, internally consistent, and temporally stable” (Campbell et al., 1996, p. 141).

The notion of individual differences in self‐concept clarity emerged from research on self‐esteem (Baumgardner, 1990; Campbell, 1990). Findings from these experiments indicate that high self‐esteem people possess a well‐articulated and more precise sense of their personal characteristics than low self‐esteem individuals. For example, high self‐esteem participants respond more quickly than low self‐esteem participants when asked if a trait adjective describes them (Baumgardner, 1990; Campbell, 1990). In addition, when compared to those low in self‐esteem, high self‐esteem individuals report more confidence in their response to self‐description questions, are more likely to endorse extreme responses to self‐rating items, show more stability in their self‐descriptions over time, and are more consistent in their self‐descriptions (Campbell, 1990). Compared to those high in self‐esteem, people low in self‐esteem tend to use a broader range when describing how they fall on personality dimensions (Baumgardner, 1990) and are less likely to use information about themselves when making personal decisions (Setterlund & Niedenthal, 1993).

More recently, investigators have become interested in self‐concept clarity as a personality construct in its own right. For example, Campbell et al. (1996) developed a self‐report scale to assess individual differences in self‐concept clarity. The researchers produced evidence for the stability of scale scores and the construct validity of the instrument. People who score high in self‐concept clarity also tend to be high in self‐esteem and extraversion and score low on measures of neuroticism, anxiety, and depression (Campbell et al., 1996; Smith, Wethington, & Zhan, 1996). Individuals with a clear self‐concept tend to use active coping strategies, whereas individuals with an unclear self‐concept tend to use passive strategies (Smith et al., 1996).

The present set of experiments was designed to further explore the nature of individual differences in self‐concept clarity. Specifically, we propose that high self‐concept clarity people are more likely than low self‐concept clarity people to rely on self information to guide their behavior. A broad literature in personality and social psychology describes various ways individuals rely on internal standards, goals, and cognitive representations of self to regulate their actions (e.g., Bandura, 1997; Baumeister, Heatherton, & Tice, 1994; Carver & Scheier, 1982; Duval & Wicklund, 1972; Fenigstein, Scheier, & Buss, 1975; Higgins, 1987). Although descriptions of the process vary from model to model, there seems to be general agreement that one's self‐concept often plays a role in determining behavior. Moreover, the extent to which self information guides behavior varies as a function of the situation and the person. We argue that people high in self‐concept clarity are more likely than individuals low in self‐concept clarity to use self information because this information is stored in a clear, consistent, and well‐articulated manner, thus making the information highly accessible and easy to process. When faced with a decision about behavior, relevant self information stored in a clear and accessible fashion is more likely to be utilized than self information that is less clear and less accessible. For example, if I have a clear sense that I am an assertive individual, then I am likely to use that self‐concept information when deciding how to act in a situation in which assertive behavior is a reasonable option. If information about my assertiveness is less clearly stored in memory, I am less likely to use that information when deciding how to act when the assertion situation emerges.

Self‐concept clarity is relevant to social influence researchers because the research reviewed above indicates that social influence appeals may vary in success depending on a person's self‐concept clarity. Specifically, the present research addresses the question of the conditions under which an individual will accept a label or manded altercast. Altercasting occurs when an influence agent adopts certain behaviors oriented toward placing a target of influence into a role that specifies the expected social behavior (e.g., telling someone that they are altruistic, then providing them with an opportunity to donate to charity; Pratkanis, 2000; Weinstein & Deutschberger, 1963, 1964). This works because a social role provides normative feedback on expected behavior for a given social context.

The effectiveness of altercasting was demonstrated by in an experiment by Miller, Brickman, and Bolen (1975). In this study, children were given false feedback that they were very neat and tidy. Miller et al. compared this altercasting manipulation against a persuasive communication urging the students to be neat and tidy or a no message control. Results indicated that placing the children into the role of being tidy by labeling them as tidy was significantly more successful at eliciting tidy behavior as compared to the persuasion condition or the control—and this effect lasted over time. Additionally, Strenta and DeJong (1981) administered what was described as a short personality test to a group of participants. Participants in one condition were given feedback indicating they were very kind and helpful people. As these participants left the experimental setting, they encountered a confederate who appeared to accidentally drop a stack of computer cards. Participants who had received the false feedback were more likely to stop and help the confederate than participants receiving no feedback.

A recent set of experiments by Burger and Guadagno (2003) support this analysis. Participants in that research were presented with a foot‐in‐the‐door manipulation in which they were first asked to perform a small request (e.g., wear a button to promote National Cancer Awareness Month) and later presented with a larger request (e.g., volunteer 3 hours to promote cancer awareness). Differences in compliance based on condition and level of helpful‐specific self‐concept clarity were examined. Only participants high in self‐concept clarity showed the usual foot‐in‐the‐door effect, i.e., an increase in compliance to the larger request relative to a control group that did not receive the initial request. The researchers explain this finding within both an altercasting and a self‐perception theory analysis. That is, complying with the initial request caused participants to think of themselves as the kind of person who supports these kinds of causes. Or, in altercasting terms, placing someone into a helpful role made him or her more susceptible to a compliance tactic oriented toward eliciting helping. However, only the high self‐concept clarity participants relied on this temporary alteration in self‐concept when deciding whether to go along with the second request.

The present research provides evidence for individual differences in susceptibility to manded altercasting. Specifically, across three experiments, an experimenter placed participants into role of helpfulness or honesty by providing false feedback on that dimension. In each of three experiments, we presented participants with bogus personality test feedback. We predicted that high helpful and high honest self‐concept clarity individuals would rely on this information and alter their behavior in line with the false feedback more than people low in self‐concept clarity.

EXPERIMENT 1

In the first experiment we provided some participants with false feedback on a personality test that described them as particularly helpful individuals. Past experiments have found that such feedback is an effective method for temporarily changing how altruistic people believe themselves to be (Batson, Harris, McCaul, Davis, & Schmidt, 1979; Kraut, 1973; Thomas & Batson, 1981). Our methodology was adopted from the Strenta and DeJong (1981) work described above. Based on our earlier analysis and prior work in this area, we expected high self‐concept clarity individuals to respond to a similar false feedback manipulation more than those low in self‐concept clarity.

Method

Participants

A total of 100 undergraduates (39 men and 61 women) served as participants in exchange for class credit.

Procedure

Small groups of participants (three or fewer per group) were randomly pre‐assigned to either the feedback or no feedback condition. As participants arrived, the experimenter greeted them and explained that they would fill out a series of personality tests during the experimental session. The first personality test completed by all participants was a modified version of the Latitude of Self‐Description Questionnaire (Baumgardner, 1990). This 20‐item questionnaire was developed by Baumgardner (1990) to measure self‐concept clarity. Test takers rate themselves on each item (e.g., intelligent, popular) using two methods. On a 100‐point scale, participants first select the percentile point that best describes them. Next, using two arrows, participants indicate the percentile range within which they believe they fall. For instance, the instructions describe a possible response to the term athletic as follows: “if you are sure you are more athletic than at least 15% of the population, then place an arrow halfway between the 10 and 20 … and if you are sure you are not more athletic than 90% of the population, then put the second arrow there.” To calculate self‐concept clarity, the range responses for the 20 items are added together. A low score indicates high self‐concept clarity.

To assess self‐concept clarity on the dimension of helpfulness, we used a modified version of the scale developed by Burger and Guadagno (2003). Seven new items were added to the original twenty to measure self‐concept clarity for the specific trait of helpfulness: charitable, unselfish, helpful, insensitive, unsympathetic, self‐centered, and compassionate. Pilot testing on this subscale with a sample of 109 participants indicated these items were strongly correlated (rs ranged between .42 and .74) and formed an internally consistent measure, α = .90.

The experimenter waited until the participant completed the Latitude of Self‐Description Questionnaire before collecting it and administering the second scale. The second scale was identified as the Interpersonal Orientation Test. Scale instructions asked participants to indicate whether each of 40 sentences described them by answering True or False on each item. In reality, the Interpersonal Orientation Test is a bogus personality scale. The items are comprised of a set of generic sentences generally having to do with social interaction that were written by the authors for the purpose of the experiment. Examples of items from the bogus personality scale include the following: “I enjoy my friends a lot,” and “many conversations are a waste of time.”

The experimenter collected the second scale when the participant finished. Participants assigned to the no feedback condition were simply given the third test at this time. Participants assigned to the feedback condition were told that the experimenter had scoring keys for the second scale. Further, the experimenter explained that the project supervisor had given him permission to score the test and provide participants with feedback if they wanted it. All participants said they would like to receive feedback on the test. Participants in this condition worked on the third scale while the experimenter pretended to score their test. The third test was the Rotter (1966) Locus of Control Scale, which in reality was used as a filler task and not scored.

When participants in the feedback condition completed the third scale, the experimenter gave them two feedback sheets from the bogus personality test. The first sheet was labeled “The Interpersonal Orientation Test – Feedback Sheet” with the participant's name written in an appropriate space at the top. The feedback sheet began with a paragraph describing the test. Briefly, the paragraph explained that each person tends to enter into social situations with a type of “social orientation.” This social orientation consists of the goals, strategies, and expectations we have for social encounters. The paragraph further explained that there were five basic types of social orientation—Aggressive, Avoidant, Engaging, Helpful, Manipulative—and that the IOT was designed to identify which of these was the test‐taker's primary social orientation. Examples were given for the different social orientation types. The bottom half of the feedback sheet contained a space listing each of the five social orientations, with Helpful circled in red ink. This was followed by feedback indicating that the participant had received a score of 12 on the Helpful section of the test and that this placed the participant in the top 85th percentile for this type of social orientation.

The second feedback sheet was a photocopy of two paragraphs labeled The Helpful Social Orientation Person. The experimenter explained that this was the test‐maker's description of the participant, based on his or her primary social orientation type. The two paragraphs described a person who is very helpful, pleasant, empathic, trusting, and who enjoys helping people in need. The experimenter returned to his desk (presumably to get more materials for the experiment) and left the participant with nothing but the feedback sheets for about 3 minutes. This step was taken to ensure that participants read the feedback sheets. The experimenter in fact observed that all participants appeared to read the sheets carefully. At this point the experimenter announced that the experiment was over and directed participants to the project supervisor's office where they could receive participation credit and debriefing. To ensure that no one else was present when the confederate dropped the pamphlets, participants were dismissed one at a time.

The location of the experimental room was selected so that participants could walk in only one direction to reach the project supervisor's room and receive credit. All participants passed a woman sitting in a chair in the hallway about 20 feet from the experimental room. The woman was a confederate of the experimenter and was blind to experimental condition. When the participant was about 10 feet away, the confederate stood and appeared to accidentally drop a stack of 10 pamphlets from her lap onto the hallway floor. The confederate had been trained to drop the pamphlets in such a way that they did not block the participant's direct path to the doorway he or she was heading toward. That is, the pamphlets landed neither too close to the participant that not helping would have been awkward, nor too far away that helping would seem unusual. The confederate immediately stooped to pick up the pamphlets and did not make eye contact with the participant or ask for help. The confederate recorded whether or not participants helped her with the pamphlets. A participant was identified as helping if he or she picked up at least one pamphlet.

Results and discussion

Range scores were totaled for each of the seven helpfulness‐related items on the Latitude of Self‐Description Questionnaire to create a reliable helpful self‐concept clarity score for helpfulness, α = .91, M = 163.73, SD = 79.64. The main dependent measure, whether or not the participant assisted the confederate, was recorded as a 0 to indicate no help was provided and 1 to indicate that the participant did help the confederate.

We selected participants with helpful scores in the upper and lower 40% of the distribution for the analysis.1 These participants comprised the high and low helpful self‐concept clarity conditions. We first compared the number of participants who helped the confederate in each condition within a 2 (condition) by 2 (high–low self‐concept clarity) loglinear analysis. The interaction effect fell short of statistical significance in this analysis, z = 0.95. We then compared the helping behavior of high and low helpful self‐concept clarity participants with two planned cell comparisons. As shown in Figure 1, participants low in helpful self‐concept clarity did not vary their amount of helping as a function of the feedback, χ²(1, N = 48) = 0.29. However, participants high in helpful self‐concept clarity helped considerably more when receiving feedback than when receiving no feedback, χ²(1, N = 52) = 5.19, p = .03, φ = .32.

We also calculated the average helpfulness score for each participant by summing and averaging the participant's ratings on the seven helpfulness items, α = .89, M = 63.46, SD = 8.88. The correlation between helpful mean scores and helpful range scores was marginally significant, r = −.176, p = .08. This indicates that participants who had a clear self‐concept tended to see themselves as higher in helpfulness than did those participants who had less clear helpful self‐concepts. Consistent with this correlation, high helpful self‐concept clarity participants had a higher average helpfulness score (M = 65.85) than the low helpful self‐concept clarity participant (M = 61.08), t(87) = 2.49, p = .02, d = 0.54. However, when we entered the mean helpfulness score as a covariate into the loglinear analysis, the interaction effect did not change appreciably, z = 0.90. Thus, although the two groups differed in their average level of self‐rated helpfulness, this difference cannot account for the different reactions to the feedback by the high and low self‐concept clarity participants.

Figure 1. Percent of participants who helped the confederate as a function of self‐clarity of helpfulness and feedback condition.

To further examine self‐rated helpfulness as an explanation for the results of Experiment 1, we additionally ran the same 2 by 2 loglinear analysis above with self‐rated helpfulness instead of helpful self‐concept clarity. As with the analysis with self‐concept clarity, the interaction of feedback condition and helpful self‐concept fell short of statistical significance in this analysis, z = −1.02. We then compared the helping behavior of high and low helpful self‐concept participants with two planned cell comparisons. As with helpful self‐concept clarity, participants low in helpful self‐ratings did not vary their amount of helping as a function of the feedback, χ²(1, N = 49) = 0.11. However, participants high in helpful self‐ratings helped considerably more when receiving feedback than when receiving no feedback (62% vs 42%), χ²(1, N = 51) = 6.33, p = .012, φ = .35. These results are not surprising given that the helpful self‐ratings and the helpful self‐concept clarity scores were correlated such that individuals with clear helpful self‐concepts also tended to rate themselves as more helpful. In fact, the high helpful self‐concept clarity and high helpful self‐ratings participants who received feedback that they were helpful were by far the most helpful group, with 94% of participants in this condition helping the confederate.

Overall, the findings from Experiment 1 are consistent with our hypothesis. When faced with the dropped pamphlet situation, the high helpful self‐concept clarity participants appeared to use the personality test feedback information when deciding whether to help the confederate. Because that information said they were helpful, these participants were more likely to stop and help than when they were not presented with the false feedback. Because low helpful self‐concept clarity participants are less likely to rely on such self information to guide their behavior, these participants did not alter their behavior as a result of the test feedback.

EXPERIMENT 2

The second experiment was designed to demonstrate not only that the effect uncovered in Experiment 1 is replicable, but also that it is not limited to some unique aspect of the trait or behavior we examined in the first investigation. We again used the false‐feedback procedure, but modified the test and materials to indicate to some participants that they had scored high on an honesty scale. Shortly after the feedback, we provided participants with an opportunity to act in an honest manner. Based on the earlier findings, we expected the high honesty self‐concept clarity participants who received the false feedback to act honestly more often than those not receiving the feedback. However, we did not expect the low honesty self‐concept clarity participants to alter their behavior significantly after receiving the false feedback.

Method

Participants

A total of 105 undergraduates (30 men and 75 women) participated in the experiment in exchange for class credit.

Procedure

The procedures were similar to those used in the first experiment, with the following exceptions. First, the students participated in the experiment one at a time. Upon arrival the experimenter explained that the experiment should take about 20 minutes and therefore was worth one participation credit. This information also was on the sign‐up sheet. Second, the seven helpfulness items on the modified Latitude of Self‐Description Questionnaire were replaced with seven items related to honesty. The new items were truthful, devious, honest, forthright, deceitful, ethical, and honorable. The internal consistency correlation for the mean scores indicates that the items appear to be measuring the same concept, α = .81, M = 72.50, SD = 11.4.

Third, the bogus Interpersonal Orientation Test was modified to include several items related to honesty. The name of the test was changed from Interpersonal Orientation Test to the Personal Attributes Test. Fourth, participants in the feedback condition were shown a test feedback sheet indicating their “primary personal attribute” was Honesty (rather than Achievement, Approval, Attachment, or Self‐Protection). Participants read that they had scored in the 94th percentile on the Honesty subscale, and read a short description of the prototypic person who has this primary personal attribute. The description emphasized that being honest was very important to someone with this kind of test score.

Fifth, upon completing the third personality test the experimenter excused himself or herself from the room for approximately 2 minutes to allow the feedback participants to read the feedback sheets. Upon his or her return, the experimenter told the participant the experiment was over, and directed him or her to a nearby room to receive participation credit. Sixth, the participant arrived at the second room to find another experimenter who was blind to feedback condition and, of course, the participant's self‐concept clarity level. The second experimenter asked which experiment the student had participated in. All participants gave the name of the experiment or a short description. The second experimenter said, “I think that's the study that takes about an hour and a half to complete, so you are supposed to get two research credits.” The experimenter then paused for approximately 3 seconds while presumably looking for a pen. If the participant corrected the experimenter during this pause, the experimenter said, “Oh, right. You were in the one‐credit study.” If the participant did not correct the experimenter during the pause, the experimenter said, “Oh, wait. My mistake. You were in the one‐credit study. Sorry.”

Participants' responses were coded by the second experimenter as “honest” if the participant either corrected him or her (for example, “No, I think it was only one credit,” “It only lasted about twenty minutes”) or indicated that he or she was unsure about the two credits (for example, “I could go back and check with the experimenter to see.”). A response was coded as “not honest” if the participant said nothing or agreed that he or she should receive two credits. Most of the “not honest” responders simply failed to say anything about the experimenter's obvious mistake.

Results and discussion

Participants' range scores for the seven honesty items were summed to obtain a self‐concept clarity score, α = .93, M = 171.82, SD = 93.31. The correlation between honest mean scores and honest range scores was significant, r = −.257, p<.01. This indicated that, similarly to Experiment 1, participants who had clear honestly self‐concepts also saw themselves as honest. The main dependent measure, whether or not the participant reported the mistake, was recorded as a 0 to indicate that the participant did not honestly report the mistake and 1 to indicate that the participant did report the mistake.

Those falling at or above the 60th percentile of the distribution were identified as low honesty self‐concept clarity participants and those falling at or below the 40th percentile were identified as the high honesty self‐concept clarity participants. The number of participants who gave honest responses in each condition was compared within a 2 (condition) by 2 (high–low honesty self‐concept clarity) loglinear analysis. The interaction effect in this analysis fell short of statistical significance, z = 1.73, p<.09. We then compared high and low honesty self‐concept clarity participants within two planned comparisons. As shown in Figure 2, high self‐concept clarity participants were significantly more likely to give an honest response when they received the false feedback than when they did not, χ²(1, N = 55) = 4.73, p = .03, φ = .29. In contrast, the low honestly self‐concept clarity participants in the feedback and no‐feedback conditions did not differ significantly in their honesty responses, χ²(1, N = 50) = 0.96. The findings thus replicate the effect demonstrated in the first experiment. High honestly self‐concept clarity participants again were more likely to behave in a manner consistent with the false feedback than were those low on this trait.

Figure 2. Percent of participants giving an honest response as a function of self‐clarity of honesty and feedback condition.

As in Experiment 1, we also compared high and low honesty self‐concept clarity participants on their average honesty rating by summing and averaging the ratings participants provided for the seven honesty items. Like the helpfulness ratings in the first experiment, we found high honesty self‐concept clarity participants (M = 75.09) rated themselves as more honest than low honesty self‐concept clarity participants (M = 69.48), t(96) = 2.60, p = .01, d = 0.52. However, again when we entered the mean honesty score as a covariate into the loglinear analysis we found no appreciable change in the interaction effect, z = 1.86.

To further examine self‐rated honesty as an explanation for the results of Experiment 2, we additionally ran the same 2 by 2 loglinear analysis above with self‐rated honesty instead of honesty self‐concept clarity. As with the analysis with self‐concept clarity, the interaction of feedback condition and honest self‐concept was not statistically significant, z = −1.24. We then compared the honestly of high and low honesty self‐concept participants' with two planned cell comparisons. We did not find any significant differences between honesty self‐concept participants. Specifically, contrary to the findings with honest self‐concept clarity, participants low in honesty self‐ratings varied their amount of helping as a function of the feedback (61% control vs 85% feedback), χ²(1, N = 53) = 5.53, p = .019, φ = .323. Also contrary to the results of Experiment 1, there was no significant difference for participants high in honesty self‐ratings (76% control vs 78% feedback), χ²(1, N = 53) = 0.16. Thus, it appears that these results are contrary to our findings in Experiment 1 and further support our rationale that self‐concept clarity is moderating the response to the altercasting manipulation. Finally, as in Experiment 1, when we examined honest responding broken down by condition, level of honest self‐concept, and honesty self‐concept clarity, the high honesty self‐concept clarity and high honesty self‐ratings participants who received feedback that they were honest were the most honest group, with 93% of participants in this condition being honest to the experimenter.

EXPERIMENT 3

The consistent pattern of findings in the first two experiments demonstrates that high and low helpful and honest self‐concept clarity people differ in their reaction to feedback about themselves. Experiment 3 was designed to provide one additional piece of information to help us better understand this effect. Specifically, it would be useful to know the extent to which participants thought of themselves as helpful just prior to encountering the confederate who dropped the pamphlets, or as honest just before receiving the extra participation credit. We did not include such a step in the procedures because simply asking people at that point to report how helpful or honest they believe themselves to be could have affected participants' subsequent behavior. However, because high helpful self‐concept clarity participants have clear and highly accessible self‐concepts, we believe they would have rated themselves as more helpful or more honest after the feedback relative to high helpful self‐concept clarity participants not receiving feedback or low self‐concept clarity participants who did. Experiment 3 was designed to test this assumption. We replicated the first experiment, but instead of examining helping behavior after the false feedback, we simply asked people how helpful they believed themselves to be. We predicted that high helpful self‐concept clarity participants would describe themselves as more helpful after receiving the false feedback than when they received no feedback. However, we did not expect this difference, or as large a difference, among the low helpful self‐concept clarity participants.

Method

Participants

A total of 100 undergraduates (39 men and 61 women) participated in the experiment in exchange for class credit. All participants had completed the modified Latitude of Self‐Description Questionnaire used in Experiment 1 two to three weeks earlier as part of a larger package of tests.

Procedure

As in Experiment 1, participants were randomly assigned to either the feedback or no feedback condition. The procedures were identical to those used in the first experiment except for the order of the three tests and the absence of the dropped pamphlet part of the script. Participants completed the tests in the following order: the bogus Interpersonal Orientation Test, the Rotter Locus of Control Scale (time filler), and a modified version of the Latitude of Self‐Description Questionnaire. As in the first experiment, feedback participants had approximately 3 minutes to examine the false feedback sheets given to them after completing the Rotter scale. The modified Latitude of Self‐Description Questionnaire contained the seven helpfulness items used in the first experiment. However, participants were asked only to rate themselves on each 100‐point scale and did not provide information about their range of scores.

Results and discussion

The range scores for the seven “helpful” items on the Latitude of Self‐Description Questionnaire completed prior to the experiment were summed to determine the participant's self‐concept clarity score, α = .93, M = 156.75, SD = 98.99. The mean scores for the seven helpfulness items on the pre‐ and post‐manipulation Latitude of Self‐Description Questionnaire were summed (after correcting for direction) to create pre‐ and post‐manipulation helpfulness scores α = .86, M = 68.98, SD = 11.49. The correlation between helpful mean scores and helpful range scores was significant, r = −.276, p<.01. High helpful self‐concept clarity participants did not rate themselves as more helpful on the pre‐manipulation measure than did the low helpful self‐concept clarity participants (Ms = 49.20 and 50.05, respectively), t(99) = 0.53.

Participants' post‐manipulation helpfulness scores were examined within a 2 (high–low helpful self‐concept clarity) by 2 (feedback–no feedback) ANCOVA, with pre‐manipulation helpfulness scores entered as a covariate. As shown in Figure 3, a significant main effect was found for feedback condition, F(1, 96) = 6.29, p = .02. However, as seen in the figure, this effect is best understood within the two‐way interaction, F(1, 96) = 3.81, p = .05. Specific planned comparisons confirm that high helpful self‐concept clarity participants rated themselves significantly more helpful after receiving the false feedback than they did when not receiving the feedback, t(49) = 2.02, p = .05, d = 0.46. However, the feedback manipulation had virtually no effect on the low self‐concept clarity participants, t(49) = 0.034.

Figure 3. Self‐ratings of helpfulness on post‐feedback questionnaire as a function of self‐clarity of helpfulness and feedback condition.

The findings thus support our interpretation of the effect demonstrated in the first two experiments. High helpful and honest self‐concept clarity participants in those investigations altered their behavior to be either more helpful or more honest after receiving personality test feedback indicating they were especially helpful or honest. We argue that these participants used self information when making their decisions to help or act honestly. That is, there are many reasons why a person might help or not help the confederate who dropped the pamphlets or choose to correct someone who had mistakenly given them too much. One of these is the extent to which the individual thinks of himself or herself as a helpful person. Because such information is stored in a clear and accessible manner for high helpful and honest self‐concept clarity people, these participants not only thought of themselves in line with the test feedback, but they appeared to use that information to guide their decision to help.

GENERAL DISCUSSION

The results from the three experiments clearly demonstrate that high and low helpful and honest self‐concept clarity individuals differ in the extent to which information about the self guides their behavior. To illustrate this effect, we used a false‐feedback procedure, which is a type of altercasting. In each experiment high helpful or honest self‐concept clarity participants responded to the personality test feedback by being more helpful (Experiment 1), or acting more honestly (Experiment 2), or thinking of themselves as more helpful (Experiment 3) than when not receiving the feedback. However, in each case feedback participants low in helpful or honest self‐concept clarity did not differ from those not receiving the false feedback. The findings are consistent with the description of self‐concept clarity. People high in this individual difference have clear, well‐articulated representations of their self‐concepts. Because self information is stored in memory in a clear and organized manner, this information may be more accessible for high self‐concept clarity people than for low self‐concept clarity people. Because self information is less clear and less accessible for people low in self‐concept clarity, these individuals are likely to rely on information from other sources when making decisions about how to act.

We do not mean to suggest that the false personality test feedback permanently altered the way our participants thought of themselves. It may have been the case that the feedback merely primed relevant information about helpfulness or honesty already in the self‐concepts of the high self‐concept clarity participants much in the same way that the foot‐in‐the door manipulation made consistency of behavior salient in our prior research (Burger & Guadagno, 2003). For instance, it may be that these results may due to priming (see Bargh, 2006, for a review). Specifically, it may be that having participants fill out an assessment of self‐perceived helpfulness or honesty followed by the labeling participants as helpful and honest primed them for honest responding. To rule out this explanation, future research should include a condition where participants receive feedback that they are not high on the relevant dimension (e.g., not a helpful person or not an honest person) to disentangle this issue. However, within the scope of the current data, we do not think that there is strong evidence for a priming interpretation because we have no reason to predict that individuals high and low in self‐concept clarity would respond differently to priming. Nonetheless, our data indicate that the high helpful and honest self‐concept clarity participants relied on this information when deciding how to act, but the low helpful and honest self‐concept clarity participants did not.

The results of these experiments have important implications for social influence researchers. Specifically, we have identified an individual difference characteristic that moderates susceptibility to an influence appeal. In this case, individuals who were low in self‐concept clarity on dimensions of helpfulness and honesty were unresponsive to an attempt to altercast them as helpful or honest. Individuals who were high in self‐concept clarity on these dimensions were responsive to the manded altercasting manipulation and were significantly more likely to respond in a manner consistent with the attribute with which they had been labeled. This is consistent with the findings of Burger and Guadagno (2003) that, compared to individuals low in helpful‐specific self‐concept clarity, individuals high in helpful‐specific self‐concept clarity were more susceptible to compliance via the foot‐in‐the‐door technique. Future research should continue to examine the impact of individual difference characteristics on influence appeals, as this can help explain some of the variance in the outcomes of social influence attempts.

Although the findings from the three experiments advance our understanding of self‐concept clarity and social influence processes, several important issues remain. We will touch briefly on three of these. First, because research on self‐concept clarity is still relatively new, there are questions about how to measure this construct. We limited our assessment of self‐concept clarity to a modified version of Baumgardner's procedure. This strategy has the advantage of allowing for easy comparison across experiments. However, other methods for measuring self‐concept clarity are available. These include a self‐report inventory developed by Campbell and her colleagues (Campbell et al., 1996). The relation between the different measures, and whether results similar to the ones reported here can be replicated with different instruments, also remain open questions.

Second, as is the case with all individual differences, other personality variables are no doubt correlated with self‐concept clarity. For example, we expect that self‐concept clarity may be correlated with self‐esteem, self‐monitoring, self‐esteem stability, attributional style, etc. As such, the possibility that other personality traits played a role in our results cannot be ruled out. However, we argue that at a conceptual level the extent to which a person's self‐concept is clear and well articulated should have an effect on how much that self‐concept guides behavior. Teasing apart the effects of other personality traits within this experimental procedure remains a goal for future investigations.

Third, there is one alternative explanation we must acknowledge. The results of this study may be due to cognitive dissonance (see Festinger, 1957, for a review). Aronson (1992) asserts that most individuals strive to be consistent, competent, and moral. As a result of these motivations, people reduce dissonance when they are put in a position where it appears that they are hypocrites because their thoughts and actions are not consistent, competent, or moral when compared to prior thoughts and actions. It follows that people will adjust their beliefs to reduce dissonance when they are put in a position where it appears that they are being hypocritical.

This perspective on dissonance ties it directly to self‐concept. Aronson, Friend, and Stone (1991) found that individuals who videotaped a pro‐condom‐use message and were then asked to recall past experiences where they had not used a condom reduced dissonance by increasing the strength of their intended future condom use. This allowed them to avoid appearing to be a hypocrite. In the present investigation, we found that the people who were high in helpful‐ and honest‐specific self‐concept clarity and were labeled helpful or honest displayed the most helpful and honest behavior. Because these participants had a very clear self‐concept on the domain in question, behaving in a manner that was inconsistent with the label would have led them to experience dissonance. By helping and being honest after receiving these labels, high self‐concept clarity participants were able to reduce dissonance. Participants who had low levels of helpful and honest self‐clarity may not have experienced any dissonance, as they had no clear self‐concept to be consistent with. These results only appeared in the feedback condition because there were no set labels for the high self‐clarity individuals to be consistent with in the no feedback condition. Thus, it may be that one of the mechanisms behind our findings that self‐concept clarity moderates response to false feedback is dissonance. Future research should examine this question further.

Notes

1. We selected participants with scores in the upper and lower 40% of the distribution for the analysis. There were a number of reasons why we chose to look at extreme scores rather than the full distribution of self‐concept clarity scores. First, the distribution of scores on this scale indicated that many participants' scores fell in the center of the distribution. One reason we chose to leave them out is that previous research has indicated that moderate scorers can obscure or misrepresent actual differences between individuals scoring high and low on a scale (Sorrentino & Short, 1977). In addition, since our dependent variable was nominal, we chose to use extreme groups to increase our power to detect a significant effect. Other researchers examining phenomena using nominal data have also used this approach (see, for example, Burger & Guadagno, 2003; Cialdini, Trost, & Newsom, 1995).