self-evaluation maintenance theory

n. The self-evaluation maintenance (SEM) model attempts to capture some of the dynamics underlying our reaction to the performance of others, particularly close others. Sometimes the outstanding performances of close others make us feel good. Albert beams with pride when he tells us that his friend Bob is selected as the first chair in the school orchestra. Other times the outstanding performance of close others can be quite negative. In spite of Albert's forced smile, he recognizes some very negative feelings in himself when he learns that his friend Charlie made the starting football team. In both cases Albert was outperformed by a friend, yet his responses to those performances were totally opposite in character. How might we understand this?

The SEM model starts with the assumption that people want to feel good about themselves: that is, they want to maintain a positive self-evaluation. The model further assumes that the performance of others, particularly close others, is consequential to self-evaluation. The self- relevance of a performance domain is also an important determinant of our response to another's performance. For each of us, it is important to be successful in a few particular areas but not so important to be particularly good in many other areas. In a word, some performance domains are more relevant to our self-definition than are other areas. For example, Albert plays the piano but does not think of himself as a musician. Music is low in relevance for him. On the other hand, he cares very much about his performance on the football field. Football is high in relevance.

The outstanding performance of others affects us through two separate processes: the comparison process and the reflection process. The more self-relevant is a particular domain the more important is the comparison process (relative to the reflection process). When a friend or relative performs better in a highly self-relevant domain, then self-evaluation is likely to suffer by comparison. Such threat via comparison will produce negative feelings. On the other hand, if the performance domain is low in self-relevance, then the reflection process is likely to be more important (relative to the comparison process). The outstanding performance of a friend or relative in a low-relevance domain can boost self-evaluation via “basking in reflected glory.” Such basking will produce positive feelings.

Research on the SEM model. Research on the SEM model focuses on three variables that can be measured and/or manipulated: (1) another's performance relative to the self. The effects we noted make sense only if the other outperforms the self. One will not suffer by comparison if one is not outperformed. And there is little to be gained by basking in the reflected glory of someone whose performance is mediocre. A second variable is closeness. Closeness refers to the extent to which self and other are “connected” to one another. The connection can be affective, as friends are closer than strangers; it can be genetic, as relatives are closer than strangers. The connection can even result from the changing context. Two women, even strangers, will be closer to one another in a class with 20 men than in a class with 20 other women. Two people from Asheville, North Carolina, will be closer to one another in Dallas, Texas, than in Asheville. Closeness is important because it intensifies the comparison and reflection experience. For example, we are less likely to compare ourselves to people with whom we have little or no connection, and the results of such comparison are likely to be less extreme. The same is true for reflection. One can almost see the pride and feel the reflected joy when the mother introduces her son: “Meet my son the doctor.” All of those feelings are absent when the mother introduces a person to whom she has little connection: “Meet John Smith the doctor.” The third variable that is either measured or manipulated in SEM research is the relevance of the performance domain to one's own self-definition. As noted, relevance is crucial because it determines the relative importance of the reflection and comparison process. When the performance domain is highly relevant, then the good performance of a close other is likely to result in comparison and a threat to self-evaluation; when the performance domain is low in personal relevance, then the good performance of a close other is likely to result in reflection and a boost to self-evaluation.

Predictions. The underlying assumption of the SEM model is that people behave in ways that tend to maintain a positive self-evaluation. This assumption allows predictions regarding performance, closeness, and relevance. If any two of these variables are fixed, then one can make unambiguous predictions regarding behavior that will impact the third variable. Let us focus on what happens to each of the three variables in turn when the other two variables are fixed.

Often we can affect the performance of another person. We can facilitate his/her performance by helping, or we can hinder his/her performance by creating difficulties that he/she must overcome. When will we help and when will we hinder the performance of another? The SEM model suggests that the answer depends on the relevance of the performance domain and the closeness of the other: When relevance is high, self will suffer via comparison to another who performs well, particularly if that other person is psychologically close. If we wish to maintain a positive self-evaluation, then, perhaps surprisingly, when relevance is high, we may be less motivated to help a close other, such as a friend, than a distant other, such as a stranger. On the other hand, low relevance creates the opportunity to bask in the reflected glory of another's good performance, particularly if that other is close. This leads to the expectation that when relevance is low, we will be more motivated to enhance the performance of a close other than the performance of a distant other.

An experiment conducted by Tesser and Smith tested these predictions. Each participant was required to take a friend to the laboratory and each session included two friendship pairs. The pairs were strangers to one another. Each participant, in turn, was required to guess target words on the basis of clues given by the other three players. Each of the three other players anonymously selected clues from a list of clues graded for difficulty. The experimenter selected from among these clues on each round of the task. Half of the participants were in a high-relevance condition. They were led to believe that performance on the task was related to such traits as verbal ability and intelligence; participants in the low-relevance condition were told that is spite of the verbal nature of the game, research had shown that it was unrelated to such traits as verbal ability and intelligence. In sum, each participant had an opportunity to affect the performance of a friend and of a stranger on a task that was either high or low in relevance. If he/she wanted to facilitate the other's performance, he/she could give clues that made it easy to guess the target word; if he/she wanted to hinder the other's performance, he/she could select difficult clues. The results were just as predicted by the SEM model. When the task was highly relevant, participants were more helpful (gave easier clues) to a stranger than they were to their friend; when the task was low in relevance, people were more helpful to their friend than to the stranger.

What about closeness? When will we want to be more connected to another? When will we try to cut the ties that connect us? The SEM model predicts that the answers to these questions depend on the state of the other two variables, relevance and performance. When relevance is high, the comparison process is prepotent. A better-performing other is threatening. Since closeness intensifies this threat, the better another's performance, the greater the motivation to reduce closeness. When relevance is low, the reflection process is prepotent. When another outperforms us, there is an opportunity for reflection, particularly when the other is close. Thus, the model predicts that when relevance is low, the better the other's performance the greater the motivation to increase closeness.

This prediction was tested by Pleban and Tesser. There were two participants in each session; a real participant and a participant who worked for the experimenter posing as a real participant. At the beginning of the session participants received a list of topics such as movies, current events, hunting and fishing, and American history. For each topic they independently and privately indicated how knowledgeable they were and the extent to which it was important for them to be particularly knowledgeable. These ratings served as an index of the personal relevance of each domain. The two participants then participated in a “college bowl” quiz on one of the topics. For half of the real participants, the topic was one that they indicated was highly relevant; for the others, it was a topic that was low in relevance. The experimenter asked topical questions. The speed and correctness of the answers by each participant determined his/her score. By prearrangement, the person posing as a participant outperformed some of the participants and did not outperform the others. Finally, participants went into a second room to fill out additional questionnaires. The posed participant always entered the second room first and always sat in the same seat. The experimenter surreptitiously literally measured how close the real participant sat to the posed participant. The SEM prediction was confirmed. When the college bowl quiz was on a high-relevance topic, the better the performance of the posed participant, the farther away from the posed participant the real participant sat. When the college bowl quiz was on a topic of low relevance, the better the posed participant's performance, the closer to the posed participant the real participant sat.

Relevance. Can the performance of another affect the way we think about ourselves? Can it increase or decrease the importance of a performance domain to our own self-definition? The SEM model suggests that it can. If we claim a performance domain to be selfrelevant, then we are likely to suffer by the better performance of another, particularly a close other. So, the better another's performance the more we should be motivated to reduce the self-relevance of the performance domain, and this should be particularly the case when the other person is psychologically close.

Tesser and Paulhus tested this prediction. Pairs of participants were scheduled for the same session. To vary closeness, half of the participants were told that they were scheduled at the same time because previous information indicated that they were very similar to one another; the remaining participants were told that they were scheduled for the same session because they had almost nothing in common. Their task was to work on a new personality measure, a dimension called “Cognitive Perceptual Integration” or CPI (CPI is a fictitious dimension invented so that participants had no prior standing on it). The participants then worked on a computer task that purported to measure CPI. Then they were given feedback on the CPI test: half the participants were told that they outperformed the other; half were told that the other outperformed them. They then indicated the personal importance of CPI in an interview and on a questionnaire. Finally, each participant was left alone in a booth waiting. In the booth were two binders: one contained biographies of people high in CPI; the other contained biographies of people low in CPI. The experimenter surreptitiously measured the amount of time participants read about people high in CPI relative to the amount of time they read about people low in CPI. Again the SEM model predictions were validated. Participants who believed that they were outperformed by the other participant said that CPI was less important to them and spent less time reading the high-CPI biographies than participants who believed that they outperformed their partner. Importantly, this relationship was more pronounced when the partner was described as similar (close) rather than dissimilar.

SEM and emotions. The SEM model does a powerful job of predicting overt behavior. Often these overt changes in behavior are associated with recognizable emotions. For example, Tesser and Collins found that when people are asked to recall their feelings when outperformed by another, they recall anger, disgust, envy, frustration, jealousy, sadness, and shame. This is particularly the case when the performance domain is relevant to the self. On the other hand, outperforming another leads to the recollection of emotions like happiness, hope, and pride. Again, this is particularly the case when the performance domain is relevant to the self.

Sometimes the feelings that we experience in connection with a close other's outstanding performance are not clearly recognized. Indeed they may not even be consciously processed. For example, Tesser, Millar, and Moore videotaped the faces of people given feedback regarding the better or poorer performance (compared to the self) of either a friend or a stranger on a task that was either high or low in relevance. Participants rated their mood in connection with these patterns of feedback. Videotapes of the face were also taken because positive and negative feelings can be distinguished in facial expressions and because such feelings may be available for viewing even when they are fleeting and even when the person is unaware of them. The findings from the study are too complicated to describe completely here. However, the point to be made is that the emotions rated from the facial displays more closely followed even the subtle predictions of the SEM model than did the self-rated emotions.

Why is there a disconnect between what the body indicates and verbal self-reports? Self-reports are more vulnerable to at least two biases: social desirability and “naïve theories.” For example, it may be socially undesirable to admit or even to believe that one feels bad upon learning that a friend did well. Such reports may be suppressed or perhaps even repressed. Also, since feelings in real situations are often complex, fleeting and unrecognizable, our naïve theories of how we are likely to feel can guide our verbal responses. For example, we might think that it is likely that we experience positive emotions when we do well and negative emotions when we do poorly; positive emotions when friends do well and negative emotions when they do poorly. Such theories have a grain of truth, but they lack the subtlety of prediction embodied in a formal model such as the SEM model.

The extended SEM model. The extended SEM model was developed by Beach and Tesser in order to recognize the impact of committed relationships on SEM dynamics. The original SEM model focused on the actor's concern with maintaining his or her own self-evaluation. However, people in committed relationships must be concerned with maintaining the relationship as well. And one's partner is subject to the same SEM dynamics as the self. Outperforming a committed partner in a domain that is highly selfrelevant may be personally satisfying, but if the domain is also relevant to the partner, it can threaten the partner and, ultimately, the relationship. Thus, the joy associated with doing well on a relevant activity will be muted by the potential threat to one's partner and to the relationship. On the other hand, outperforming a partner on a task that has little self-relevance to the partner provides the partner with the opportunity to bask. In this case, both members of the dyad can take joy in the outcome.

This analysis has at least two important implications. First, in committed relationships one's response to a performance inequality will be determined not only by the relevance to the self but also by the relevance to one's partner. Second, if people are motivated to maintain their own self-evaluation and to help protect the self-evaluation of their partner, then we would expect to find a complementary distribution of relevance and performance in committed relationships. In domains where there are consistent, noticeable differences in performance, relevance should be high for the better-performing partner; relevance should be low for the poorer-performing partner. Such a distribution provides for positive affective outcomes for both partners via comparison and via reflection.

Several empirical studies using married couples have confirmed both expectations. (1) When marital partners perform on the same task, then the affective response of each partner is determined by own relative performance, relevance to self, and relevance to partner. (2) The kind of complementary distribution of performance and relevance described also tends to emerge between marital partners. Moreover, the greater the reported marital satisfaction the more pronounced is this complementary distribution. – AT