The Study of Self Esteem
Updated Ap3, 2003.
Cognition and Emotion? The Dead End in Self-Esteem Research
Thomas J. Scheff and David Fearon, Jr.
Abstract: This article suggests that studies of self-esteem using scales have reached a dead end, and suggest alternative directions. First we show how significance tests have obscured meager results. According to reviews, this huge body of research has yielded no substantial findings. Some sub-fields show consistent, but trivially small, effects; most show none at all. Most important, the size of effects does not seem to be increasing. Three questions are raised: 1. Are new standards needed to determine when to continue or stop a given line of research? 2. Are new approaches needed alternative to standardized scales and statistical tests? 3. Should future studies of self-esteem emphasize feeling and social components at least as much as cognitive (self-evaluative) components? Studies of self-esteem using interview techniques by George Brown and his colleagues suggest the need to move closer to actual data. An exploratory study that takes this direction a step further is described. We analyze social and cognitive/emotional elements second by second in discourse about topics relevant to self-appraisal and self-feeling.
Statistical tests of significance for sampling error have become the primary method in the social and behavioral sciences for evaluating research results. For almost as long, this method has been controversial, to say the least. The debate on whether and how the method should be used goes back as far as Ronald Fisher, and forward into the present. A long line of distinguished commentators (Selvin, Meehl, Bakan, Rosenthal, Cohen, etc.) have condemned the practice when used in the absence of measures that directly show effect size. All of these commentators have demonstrated that from a statistical and scientific point of view, there is no possible justification. Cohen (1994), for example, has stated that it is meaningless ritual that is impeding the advance of research in psychology.
But these warnings have not stopped or even reduced the practice. Recently a task force established by the APA published a warning against the use of tests of significance without effect sizes (Wilkerson 1999). Wilkerson goes on to note that earlier warnings have gone unheeded:
(1994) publication manual included an important new "encouragement" (p. 18) to report effect sizes. Unfortunately, empirical studies of various journals indicate that the effect size of this encouragement has been negligible (3 citations: one in 1996, two in 1998). (Wilkerson 1999, p. 599)
There are probably many reasons that the practice continues. One reason that we will pursue here concerns the nature of the criticisms. Though they have a solid foundation in statistics and philosophy of science, they are, for the most part, extremely abstract. For this reason, most social and behavioral scientists probably can’t envisage their meaning and implications, and see no reason to change their practices.
This article will take a different tack. Rather than review, once more, the scientific reasoning, it will focus on the consequences. To do this, the study of self-esteem will be used as a case study.
There have been a vast number of studies that allow us to assess how good self-esteem is as a predictor. Although most individual studies report positive results, the answer in the large is quite different. At least as self-esteem is being investigated, it seems to have either no association with external variables, or trivially small ones.
Does Self-Esteem Predict External Behavior?
Virtually all empirical studies of self-esteem use tests of significance. These studies represent an enormous expenditure of time and effort. There are more than 200 different scales that purportedly measure self-esteem. The development of so many scales alone represents an enormous effort. A still larger investment has been made in empirical research. At this point there have probably been at least fifteen thousand studies. This amount of effort probably represents the largest body of research on a single topic in the history of all of the social sciences. How has this effort paid off?
We assess four reviews that cover the total field of self-esteem scale studies (Wells and Marwell 1976;Wylie 1979; Jackson 1984; Mruk 1999) and, for comparison, five large sub-fields (occupation, social class, crime, gender, and race/ethnicity) (Chapters 9, 14, and 17, in Owens, et al 2001, Kling, et al 1999, and Twenge and Crocker 2002). We also compare our assessment with a much more comprehensive review of sub-fields by Burmeister et al (2003).
Jackson (1984) and Mruk (1999) not only sounded the alarm about the lack of consistent findings, they also reject the whole approach using standardized scales. The other seven reviewers wholeheartedly support that approach. Yet all but two of the 7 conclude that there have been no consistent results. The exceptions are a review of links with race and ethnicity (Twenge and Crocker 2002), and with gender (Kling, et al 1999) discussed below.
Wells (Ch. 14, in Owens, et al 2001) concluded that studies of the relationship between social class and self-esteem have reported findings that are "competing, inconclusive, and inconsistent." Similarly, Kaplan (Ch. 17, in Owens, et al 2001) concludes that studies of the relationship between crime and self-esteem are "rife with contradictory or weak findings." The conclusions of five of the other reviews are stated in virtually identical language.
The reviews by Twenge and Crocker (2002), and by Kling et al (1999) however, propose consistent findings. Twenge and Crocker report that the large number of studies comparing white and black subjects usually found higher self-esteem among black adults, as compared to white adults. Kling et al report that males are usually found to have higher self-esteem scores than females. But how large are these effects?
Both Twenge and Crocker and Kling et al use "d", (difference in means in terms of standard deviation) to report effect size. To get an alternative sense of the size of effect, there is an equation that relates d scores to correlation size. This equation shows that the amount of variance accounted for by the largest d (.33) in the studies reported by Twenge and Crocker is 2.7% (Table 2). Since most of the studies they reviewed of race and ethnicity reported even smaller differences, the average amount of variance accounted for in these studies is well under 2%.
The largest (d=.33) and the mean effect size (d=.21) in the Kling et al (1999) study of gender differences in 216 studies and 3 national samples seem to be remarkably similar. That is, despite the levels of significance, the effect sizes in both studies are perilously close to zero.
In a forthcoming review of findings in sub-fields of self-esteem research, Baumeister and his colleagues (2003), examine associations between self-esteem scales and external variables in great detail (at over 43 thousand words, the paper is five or six times the length of the usual article.) This review also carries weight because Baumeister may be the leading self-esteem researcher, and because the article includes a very large number of sub-fields (seventeen), ranging from academic performance to eating disorders. But their findings are similar to mine:
With the exception of the link to happiness, most of the effects are weak to modest. Self-esteem is thus not a major predictor or cause of almost anything (again with the possible exception of happiness). Moreover, the effects of self-esteem become weaker as the criteria for evidence become more objective. It is perhaps no accident that the strongest apparent benefit of self-esteem is on the most subjective outcome, namely happiness. As we noted at the outset, people with high self-esteem seem sincerely to believe they are smarter, more accomplished, more popular and likable, more attractive, and so forth, but some of those apparent advantages are illusory… (Burmeister et al 2003, p. 42).
In terms of self-esteem scale studies, their review is very comprehensive. It is also thorough, showing several ways in which findings reported as positive are in fact questionable. For example, they report that in many studies, external sources or effects of self-esteem disappear when relevant variables are controlled. The slight difference between the conclusions reached by their review and mine are only semantic. Baumeister et al summarize the results as showing only "modest," small, and null correlations. None of the studies of external prediction that they judge to show "modest" correlation report correlations over .33. Since these correlations translate to less then 10% of the variance, we would call them small, rather than modest. Although the authors are still hopeful that further research in some of the sub-fields may yet provide pay-offs, their conclusion is that support for such hope has not been found.
At present, there are no published standards for determining when a line of research should be continued or stopped. There seem to be no discussions of the matter. It’s difficult to imagine what the limits should be. For the sake of argument, let me propose an arbitrary level: if five hundred studies produce results that are only null, small, or modest, it should be discontinued. Given such a standard, it may be time that all fields of research in the social sciences be reviewed.
Perhaps there should also be standards concerning the point at which groups of studies should be reviewed as a whole. A modest criterion would be at least one per five hundred studies. By that standard, there should be at least 30 reviews, rather than just the nine we have found. And perhaps there should be at least one meta-review, like this one, per thousand studies. Then there should have been fifteen such reviews, rather than just one.
If these criteria were accepted, some lines of research in experimental social psychology (for example, studies of conformity) probably should be re-opened, since relatively few studies were conducted, yet they showed promising results. Cross-cultural studies of fundamental attribution error, since so few have been done, may be another promising field. Others, such as the present line of self-esteem studies, should be either abandoned in favor of new directions, or at least become only one of a large number of competing approaches. Perhaps there should also be discussion of the state of disciplines such as psychology and sociology in which so many studies of self-esteem continue in the face of so little results. The remaining sections of this report will consider some possible new directions in the study of self-esteem.
Tests of Significance and Dead Ends
How can so many individual studies report positive results, but lead to so few consistent or such small findings? The acceptance of findings as positive largely in terms of tests of significance seems to be the major cause. In terms of size of variance accounted for, .001 sets a very low threshold. Given large samples, .00l can occur when very little of the variance is explained, as was the case in Twenge and Crocker study discussed above. But in these cases, most of the variance is unexplained, casting doubt on the causal role of the variable being studied.
In the beginning of research in a new direction, small effect sizes mean little. But as the same direction is explored in many studies over long periods of time, they become crucial. If the approach is sound, effect sizes should be increasing. Over the many years of self-esteem research, there should have been an upward slope, no matter how slight, to the mean effect size reported in all studies. This point is the crux of our argument.
As far as we can tell in self-esteem studies, there is probably no upward slope. If a study found a large effect size, it would be reported rather than hidden. The continuing lack of direct reporting of effect size implies, it seems to me, that studying self esteem with scales has become a dead end. This is particularly true since the scales researchers have ignored the series of studies of self-esteem by George Brown and his colleagues that has shown large effect sizes (the most recent of these studies, by Andrews and Brown 1995, is discussed below).
Selvin’s critique (1957) of tests of significance was directed largely to their use with survey data. He argued persuasively that survey designs seldom meet the requirement that each of the groups compared be sampled randomly. Although his critique was widely acknowledged to be trenchant, it is our sense that it had little effect, since tests of significance are still frequently used with survey data that do not meet the randomness criterion. Like all the other critics, Selvin also urged that direct measures of size of effect always be included.
As far as we can tell, the part of Selvin’s critique dealing with randomness is probably only partially relevant to the studies of self-esteem. Most self-esteem studies use experimental, rather than survey designs. In these experiments, the separation of subjects into treatment and control groups may usually be random. But some of the studies discussed here were based on surveys (such as many of the findings in the Twenge and Crocker study). To this extent, Selvin’s critique is relevant.
But we think there are many crucial issues in addition to the randomness of sampling. One issue that needs to be discussed involves the statistical management of error. Is it better to lean towards a Type 1 error (a false negative: judging a finding to not support the hypothesis when in fact it does) or a Type 2 error (a false positive: judging a finding to support the hypothesis when in fact it doesn’t).
There are practical applications of statistical decision-making that require preference for Type 2 errors. For example, in medical diagnosis, to the extent that the suspected disease is dangerous, the diagnostician should lean toward Type 2 errors (Scheff 1963). But in scientific studies, it is generally agreed that there should be a very strong preference for Type 1 errors. The usual reasoning is that judging a hypothesis supported when it is not can be much more damaging to science than judging the hypothesis not supported when it is.
Textbooks in statistics and in philosophy of science uniformly recommend that scientific investigators direct their efforts toward rejecting, rather than supporting the hypothesis. Perhaps the underlying idea is that truth will out, even if one mistakenly judges a true hypothesis not supported. On the other hand, if one leans toward supporting one’s hypothesis, a fictitious reality can be built up. This is more or less what has happened in studies using self-esteem scales. Thousands of studies that show no or only trivially small results create an unreal world in which self-esteem, as measured by scales and significance tests, is still being investigating.
To be sure, consistent findings, even though weak, are sometimes important. One well- known example is that of Durkheim’s (1901) study of the link between suicide rates and religion. His study demonstrated that religious affiliation is consistently correlated with suicide rates. His best-known conclusion is that rates of suicide are consistently lower in Catholic areas of Europe than in Protestant areas. Subsequent studies have supported Durkheim’s findings.
Even though the amount of variance in suicide rates linked to religion is always quite small (7 % at most), these findings provided the basis for the establishment of sociology as a discipline separate from psychology. Durkheim argued that the small difference in rates demonstrated a causal factor completely independent of the psychology of individuals. This argument has been taken to mean that a discipline is needed which focuses on social, rather than psychological causation.
But even conceding that Durkheim’s findings on suicide were important historically, they and the subsequent findings like his do not establish the wisdom of settling for studies that account for only a tiny amount of variance. These findings have made no contribution to an effective theory of the causation of suicide. Some Catholics do commit suicide, and most Protestants do not. The major causative agents in suicide are still unknown. Durkheim’s speculations about the role of alienation in suicide may be insightful, but they have never been adequately tested.
Tests of significance concern only sampling error, which is usually not important compared to other types of error, such as errors of measurement, design, and conceptualization. The use of the easiest, rather than the most stringent criteria of significance is an affront to the scientific method. In real science, one attempts to find support for the null hypothesis.
The dependence on tests of sampling error, rather than the most likely sources of error, and the presentation of findings in ways that do not directly show the small effect size, give credence to Feynman’s (1985) contention that modern psychology is a "cargo cult science." It is not clear how much reading he had done in psychology. Apparently the instigation for his evaluation of psychology began with conversations he had with a psychologist whose office was across the hall. If perchance the conversations had been a sociologist instead, Feynman might have come to a similar conclusion about sociological science.
Linking Theory, Method and Data
The most likely and largest error in self-esteem studies is probably the method of measurement. The scales used for this purpose have been established to be reliable, but they are probably not valid. The standardized questions that are used in the various self-esteem scales may be tapping only the cognitive component of self-esteem, but ignoring or confounding the social and emotional components.
The bias toward cognition in self-esteem scales seems to provide the basis for Baumeister and Boden’s (1996; 1998) sensational idea that high self-esteem could cause aggression and violence. There is a thread in the self-esteem literature suggesting that high valuation of self could be defensive, beginning with Schneider and Turkat (1975). To the extent that self-esteem scales measure only cognitive self-appraisal, then an inflated evaluation of self-worth, perhaps largely a defense against painful feelings of worthlessness and/or alienation, could lead to aggression and violence.
Earlier Scheff (1994) proposed the hypothesis that destructive violence and aggression may be a defense against feelings of worthlessness. The Baumeister and Boden thesis can be re-stated as follows: self-esteem scales, because they focus on cognition rather than feeling, confound false pride with genuine pride, unacknowledged shame with the absence of chronic shame, and therefore, egotism with self-esteem.
In contrast to the massive volume of atheoretical empirical studies, theories of self-esteem are relatively few and far between. Building theory is of no value in itself, unless it leads toward research that balances theory, method, and data. It appears that human behavior is so complex that one needs link what Spinoza (Scheff 1997) called "the least parts with the greatest wholes." We interpret his idea to be that to understand the human world, we need studies that simultaneously and systematically deal with the microworld (words and gestures), and the macroworld (social institutions, general theories).
One example in the study of emotions is Retzinger’s (1991) analysis of cognitive and emotional elements in quarrels. Drawing upon social psychological theories of conflict, she developed a systematic method for locating hidden indicators of shame and anger in discourse. She applied this method to the videotaped record of four marital quarrels, analyzing exchanges second by second. On the basis of her findings, she proposed a theory of destructive conflict: unacknowledged shame gives rise to range and aggression. Of course her study was not meant to test a theoretical proposition, but only to generate one. To be able to further links between micro and macro elements relevant to self-esteem, it may be first necessary to further analyze the concept of self-esteem itself.
An Alternative Approach: Emotional and Social Components of Self-esteem
Compared to the tidal wave of empirical studies, the analytic concept of self-esteem is relatively undeveloped. The meaning of self-esteem in vernacular usage has both cognitive and affective components.
1.Pride in one’s self; self-respect. (American Heritage Dictionary. 2000)
2.Holding a good opinion of one's self; self-complacency. (Webster's Revised Unabridged dictionary.1998)
3.A feeling of pride in yourself [syn: self-pride] 2: the quality of being worthy of esteem or respect [syn: dignity, self-respect, self-regard] (Princeton University Wordline. 1997)
Note that two of the three definitions (1 and 3) offer a definition in terms of pride. Definitions # 1 and 3 also suggest other synonyms that, if not emotions, are at least mixtures of feelings and thoughts, with feelings predominating: respect, regard and esteem. These two definitions stress the affective components of self-esteem. If we assume that shame is the emotion opposite to pride, then two of the three definitions suggest that high self-esteem involves pride, and low self-esteem, shame. we return to these emotions below.
Definition # 2, however, takes a different tack: it mentions no emotions or feelings. Instead, it defines self-esteem cognitively, holding a good opinion of self. It is this second meaning that self-esteem scales appear to emulate. That is, self-esteem scales and studies focus on that part of self-concept which is verbally described and cognitive, rather than affective and social.
One aspect of the third definition also refers to yet a third component, albeit quite indirectly. "Being worthy of esteem or respect" seems to imply a social audience, perhaps in addition to being one’s own audience.
This idea is addressed by Leary and Baumeister (2000), making social acceptance a key source of self-esteem. Until this point, social acceptance has been ignored in conceptualization and research on self-esteem. They also put more emphasis on affect than is customary in the self-esteem literature. But Leary and Baumeister don’t take the next step, which is to specify the affects that they think are central to self-esteem. When they argue for the importance of the affective component in self-esteem, they approvingly cite Scheff, Retzinger and Ryan (1989), but ignore the emotions that are named in that article as specific to self-esteem, pride and shame (discussed below).
References to emotion in the abstract are virtually meaningless. Knowledge of emotions is not easily generalized, but is particular to specific emotions. For example, we know a great deal about anger. Much of what we know is probably accurate, or at least accurate enough to allow us understand each other. We have fairly specific ideas and images about the sources from which anger arises, different forms and gradations it can take, and outcomes that it can lead to. We also have similar kinds of knowledge about other primary emotions, such as fear, grief, pride, shame, contempt, disgust, love and joy.
Our knowledge about emotions held in common allows us to communicate with each other on this topic, and restrains groundless speculation. The different emotions may have several underlying similarities, but what is more obvious is the great differences in origins, appearance, and trajectories. Statements about emotions in general are extremely vague. Some of what has been described as "emotional arousal’ might appear plausible when applied to one emotion, say anger or fear, but not to most of the others. The sources, appearance and consequences of anger and fear are so different as to forbid lumping them together.
Treating all emotions together under a single heading amounts to a kind of dismissal. A current parallel can be found in rational choice theory, which divides behavior into the rational and the non-rational. In this theory, attention is given only to rational behavior. As in classical theory, the non-rational, the irrational, and emotional behavior are completely ignored.
Shame and Self-Esteem
The specific emotion of shame has been proposed as a key component of low self-esteem by Scheff et al (1989), as mentioned above, and discussed further in Scheff (1990). This idea has also been examined in Cook’s (1989) work on internalized shame, and much more extensively, in Tangney and Dearing’s analysis of shame and guilt.
Tangney and Dearing (2002, p. 32) quite rightly criticize Cook’s attempt to equate internalized shame and self-esteem. They point out that shame is an emotion, and as such, is not likely to be a continuous part of the self. Self-esteem, on the other hand, is usually assumed to be a relatively enduring aspect of the self.
They (2002, pp. 59-63) present a closely reasoned account of the relationship between shame and self-esteem. Their entire approach, however, is empirical, based on correlations between shame measures and self-esteem scales. Tangney and Dearing conclude that self-esteem is an entity that is independent of shame, but their reasoning depends entirely on the assumption that self-esteem scale measures are valid. Because of the reviews of self-esteem studies above, their conclusion may be erroneous.
Like Tangney and Dearing, the contributors to an edited volume (Baumeister 1993) on conceptualizing self-esteem seem to assume that self-esteem scales are valid, since they interpret the results of individual studies, rather than comparing the findings of groups of studies. We believe that this assumption underlies the gigantic morass of self-esteem research.
Alternative Methodologies for Future Studies
If tests of significance are too lenient in judging results positive, what would be a better method? One direction would be to require a very high proportion of the variance, say 51%. Then one would be reasonably sure that there was no other factor more important than the one under study. However, this level is impossibly high, judging from results in the field so far.
Perhaps the problem of method should be framed more broadly. The design of research in this field is almost entirely based on static situations and correlation, using standardized scales and experiments. This would be a good method to use if the purpose were to verify precise and powerful models or theories concerning the sources or effects of self-esteem. But we seem to be in a much earlier stage of investigation, since the very nature of self-esteem is still being argued. Perhaps needed at this point are studies that generate, rather than test, hypotheses. One direction would be to explore cognitive, emotional and bond-relevant responses of participants in dialogue.
The work of Andrews and Brown (1995) is a step in this direction. They simply counted positive and negative references to self in their interviews with depressed women. This method gave rise to large effects in correlations with depth of depression and with predicting improvement and relapse. This method is exploratory, but it yielded much larger effects than the typical scale-based study. This paper is only the last in a series of similar studies that Brown and his colleagues have been publishing since 1986. Yet, as far as we know, no one has undertaken a replication. Although these studies are fully cited in PsychInfo, they have not been mentioned in the subsequent reviews of self-self esteem research.
An Exploratory Study
Approaching data still more closely, Fearon and I (2003) analyzed cognitive and emotional components of two subjects’ verbal responses to all of the items from a self –esteem scale. The response of both subjects occasionally contained self-evaluations, but much more frequently there were cues to shame and to orientation to their social bonds. Indeed, these latter two aspects were dominant, being present in virtually every response.
Applying discourse analysis to two videotaped interviews that questioned subjects about their responses to a self-esteem scale, we show the sequential organization of affect occurring in response to threats to social relationships, and individual differences in the subject’s styles of managing these threats. This approach could be used to determine the validity of self-esteem measures, to complement the existing research on emphasis on reliability.
Discourse analysis grounds theory in the moment-to-moment activity of those being observed. Several methods, including conversation analysis (Heritage, 1984) and sociolinguistics (Gumpertz, 1982, Levinson, 1983, Chafe, 1980) share a close analysis of recorded social interaction. Methods typically used for self-esteem studies base theories on summaries of interaction reflected in Likert scales, and statistical or experimental analyses of populations. Through discourse analysis, we integrate theory with a direct presentation of what the subjects are doing and propose organizational features of their practices.
We define shame as painful feelings occurring on occasions of guilt, dishonor, embarrassment, condemnation, stereotyping, or remorse. Each of these situations involves social separation, or a condition of self in a troubled relation to others. Thus the term shame can be used to classify a family of terms for emotions that each refers to feelings arising from troubles in social relationships. The social sciences, however, have rarely explored the links between affects and social relationships.
The Fearon (1994) study follows from and modifies the approach used in earlier studies of pride and shame (Lewis 1971; Scheff 1990; Retzinger 1991.) These studies suggested that pride and shame are ubiquitous in social life, and are essential components to the operation of social interaction and personal identity. Shame and pride are affects dedicated to the maintenance of social bonds. These are the relationships one maintains with both real and imagined others, and one’s sense of social solidarity, or connection with a community. We will define social bonds in more detail below. Pride arises when one’s social bonds are secure.
Shame arises when people meet threats to their relationships with real or imagined others. Shame does not occur simply as a general mood or disposition accompanying a problematic social situation. Rather, shame is functionally linked to people’s detection and management of relationship trouble. People manage shame by attempting to resolve the threat to the bond.
In the Fearon (1994) study, each subject completed the 16 item Texas Social Behavior Inventory (TSBI) scale (Form A) (Helmreich and Stapp 1974). The interview portion was based on seven questions from the TSBI. This scale is not as widely employed as certain others, but is still used to operationalize self-esteem (Lammers & Becker, 1992).
Before beginning the written scale, subjects were told that they were completing a self-esteem scale, and would be interviewed using previously prepared questions about some of their responses. They were told that they would be asked to describe events from their lives that applied to the questions and should think of examples while answering the written questions. During the interview, subjects were read each question and their circled choice for "characteristic of me." They were instructed to repeat the question and their answer and briefly say whatever they wished about it. After this "spontaneous" response, they were asked the prepared questions, which encouraged them to recall concrete examples. The interviewer occasionally interjected improvised probe questions to clarify or generate details. For the most part, however, the interviewer proceeded through the question list with little interruption.
The use of spontaneous assessments and life-history examples provide two frameworks for self-evaluation. First, assessing their written response allows subjects to talk about what the question would presumably measure. They can orally assess, for example, how "people look up to me" relates to "not very characteristic of me." Secondly, the interview also provides occasion for self-evaluation as subjects manage their self-presentation as they justify their answer to the interviewer. Describing life-history situations may evoke in subjects any emotional significance of that past event, occurring with the evaluative "face-management" (Goffman 1957) in presenting their stories to the interviewer.
In this excerpt, Darryl (a first-year university student) is asked to describe a specific instance of making a decision with a group of friends. He responds with a story in which he is left to decide for his friends where to eat. The appendix lists the transcription conventions. Words in italics mark occurrences of shame. Above the text are the gestures and paralanguage features that indicate shame.
Excerpt Darryl: 13:35 - 13:47
Interviewer’s question: Describe a time when you made a decision for a group of friends.
Lowers head, rapid voice eye contact
and so: () >because of that< they thought I was like al:ways (1.1) pretty-
head down, false smile
much telling them "lets do this" and an (0.3) >getting my< wa:y
(interviewer nods,"huh" ¯) averts eyes,verbal disruption
bu:t () it wasn’t like that at all () >Theyjs wr<() >th:ey we:ren’t< () eno:ugh >
rapid voice inbreath
>they didn’t have enough to<() (•hhh) make their ow:n dec:isio:ns
1) Orienting to isolation bond threat: In lines 1 and 2, Darryl orients to an isolated social bond, in which his friends blame him for making a decision. He refers to the bond twice, with "I—telling—them" and [I]—"getting my way"—[with them]. By emphasizing the words "telling" and "getting" Darryl gives the impression that he is inflicting himself upon his friends by making the decision. The words "always" and "pretty-much" also extend the friend’s blame from the specific to general case, as though these were common judgments. His formulation of the bond suggest that his friends implicitly reject him for taking an authoritarian position. Darryl describes a fairly mild case of social rejection. What distinguishes this story as a bond threat is Darryl’s subsequent emotional response to telling the story, and his effort to reorient the bond.
2) Reorientation: In lines 3 and 4, Darryl reorients the bond in two steps. First he verbally denies the blaming situation, emphasized and extended by the words "at all." He then blames his friends. Blaming justifies his isolated position in the social bond, attributing to his friends a personal failure by which he can reject them. Reorientation is a verbal method for bypassing shame by rapidly shifting the conditions in which shame was evoked.
3) Shame: Darryl’s non-verbal behaviors during the reorientation also bypass shame. As Darryl begins to formulate the blame, he looks away, then hesitates, stammers, speaks rapidly, and draws a breath before arriving at "make their own decision." He seems to spit out the blame statement, anticipating its emotional tenor before he has cognitively formulated its linguistic content. He also hides direct reference to what his friends do not "have enough" of, in effect mitigating a more emotionally charged blame statement. Rapid speech and thought distract from any feelings arising with the orientation to being rejected, or that come with the formulation of blame.
This bond threat sequence shows Darryl’s strategy for managing the isolation bond threat posed by a negative judgment by friends, and evoked by presenting its description to the interviewer. His solution for implicit rejection by others is to justify the isolation by blaming the friends. Blaming provides a strategy for bypassing verbally the conditions in which he might otherwise feel shame. Rather than acknowledge directly, for instance, that he can sometimes act "pushy," he attributes the threat to an external source and attacks the source. This may be a common form of verbal self-defense. It is also a strategy for avoiding feelings of shame.
Darryl is able to justify an insecure bond from the position of a secure bond with the interviewer. The bond security relies on the interviewer’s implicit agreement with Darryl’s justification for blaming his friends. Darryl could feel that the interviewer finds him blameless. Notice in line 3, during the pause after "at all ()," just before Darryl blames his friends, the researcher nods and verbally acknowledges, "huh." Part of the reorientation strategy requires impression management for securing the immediate bond with the listener, while simultaneously dealing with the personal trouble of the threatened bond.
Given these criteria for maintaining bond security, bypassing shame may be more effective for managing isolated bonds than an overt display of shame. Had Darryl allowed himself overt feelings of shame, he may have felt blameworthy and guilty about blaming his friends. A display of shame, such as blushing or turning the head, would also hamper his ability to present a legitimate case to the interviewer, undermining his justification.. Darryl attempts to mitigate the blame, for example by not specifying what his friends do not "have enough" of. His strategy for managing the bond threat of isolation by rejecting the rejecters works better if he controls his shame. We can, of course, only speculate about what Darryl actually feels. The functional relationship to emotion gestures, or their absence, to the details of Darryl’s talk, however, connote the operations of underlying affect, and serve as strong empirical indicators.
Two exploratory quantitative measures were based on discourse analysis, which divided the interview transcripts into "intonation units" (IU) which are roughly equivalent to grammatical clauses. Linguistics research is finding intonation units as a reliable measure of the "rhythm of talk" by which people mark semantic and syntactic elements, which also often correspond to shifts in affect display. These units were coded for indications of shame, evaluative references to relationships, and mitigation of prior statements (reorientation) using observed gestures, linguistic details, and contextual relevance.
Of the IU’s in their interview responses, 69% of the IU’s for Daryl occurred as part of a bond threat sequence, and 85% for the female subject. Shame indicators occurred in 50% of the 597 IU’s for Daryl and 64% for the other subject. If averaged across time, there was a shame indicator every other clause, or roughly one every two seconds. Shame indicators were most often clustered in association with bond markers and clauses that mitigated prior or subsequent statements. Markov analysis of sequential probabilities supported our observations that shame occurred as part of a "bond threat sequence." Shame indicators most often followed indications of relationship troubles or negative evaluations by others, and when shame occurred, it was most often mitigated within one IU.
An estimate of the magnitude of shame based on Gottschalk et al(1969) and Gottschalk (1995) found that subjects exhibited more shame when describing scale items that indicate lower self-esteem. High self-esteem responses, however, had, on average, a slightly higher shame magnitude than would be expected for a linear correlation. Discourse analysis showed subjects often displayed shame and mitigated positive statements about themselves. These shame scores may be related to "presentation of self" in an interview context, in which there is a social stigma against presenting oneself too positively.
Our study suggests an alternative way to approach the issue of defining self-esteem. Some two hundred years ago, William Blake observed that all art and science is based on "minute particulars." Perhaps the study of self-esteem has gotten lost in large aggregates abstracted from the responses of many subjects. Can we return to the concrete particulars? With new approaches, the study of self-esteem may yet be salvaged from its graveyard of null and weak results.
This report on reviews of the field of self-esteem makes two main recommendations. 1. The conventional approach, based on scales and significance tests, should be discontinued, or at least be made to compete with alternative directions. 2. As an alternative to existing studies that are static and correlational, studies of the dynamics of discourse on topics relevant to self-appraisal and self-feeling might help new conceptions of self-esteem, and generate important and testable hypotheses.
In addition to the main recommendations, several others are implied in this review.
- When studying complex human phenomena, such as self-esteem, most of the initial effort should probably be directed toward exploration, rather than verification. Verification certainly has a role to play in science, but it usually becomes important only after considerable exploration and discussion of basic concepts and propositions.
- Similarly, atheoretical empiricism usually has a very limited application to complex human phenomena. What seems to be needed first is exploratory studies of concepts, theories, and research designs that might be relevant.
- The reliance of items in self-esteem questionnaires on self-report illustrates another continuing problem in the social and behavioral sciences. It is possible that many of the most important ingredients of self-esteem, such as feeling of adequacy/inadequacy, are not directly available in consciousness. Studies that deal only with surface variables, overt behavior and conscious content, are apt to miss the main motives of behavior. If scales are to be used, some of the items should probably be quite indirect.
- The social and behavioral sciences probably should become more reflective and self-critical. Perhaps reviews and meta-reviews of fields of study should be more frequent and thorough. We must not allow ourselves to be misled by our own PR releases. Guidelines like these might be a step toward avoiding future catastrophes like the one described here.
Andrews, Bernice; Brown, G. 1993. Self-esteem and vulnerability to depression: The concurrent validity of interview and questionnaire measures. Journal of Abnormal Psychology. Vol 102(4), pp. 565-572
Andrews, Bernice; Brown, George W. 1995. Stability and change in low self-esteem: The role of psychosocial factors. Psychological Medicine. Vol 25(1), pp. 23-31
Baumeister, Roy (Editor). 1993. Self-Esteem: The Puzzle of Low Self-Regard. New
Baumeister, Roy, Laura Smart, and Joseph Boden. 1996. Relation of Threatened Egoism to Violence and Aggression: The Dark Side of High Self-Esteem. Psychological Review. 103: 5-33.
Baumeister, Roy and Joseph Boden. 1998. Aggression and the Self: High Self-Esteem, Low Self-Control, and Ego Threat. Pp. 111-138 in Geen, R. and E. Donnerstein (Editors), Human Aggression. San Diego: Academic Press.
Baumeister, Roy F. Jennifer D. Campbell, Joachim I. Krueger, and Kathleen D.Vohs. 2003. Does High Self-Esteem Cause Better Performance, Interpersonal Success, Happiness, or Healthier Lifestyles? Forthcoming in Psychological Science in the Public Interest.
Brown, G W; Bifulco, Antonia T; Andrews, B. 1986. Social support, self-esteem and depression. Psychological Medicine. Vol 16(4), pp. 813-831
Brown, G, Bifulco, Antonia T; Veiel, H O; Andrews, B. 1990. Self-esteem and depression: I. Measurement issues and prediction of onset, pp. 200-209.
II. Social correlates of self-esteem. pp. 225-234
III. Aetiological issues. pp. 235-243
IV. Effect on course and recovery. pp. 244-249
Social Psychiatry & Psychiatric Epidemiology. Vol 25(5), Sep 1990.
Brown, G., Bifulco, Antonia T; Veiel, H O; Andrews, B. 1991. Self-esteem and depression: II. Social correlates of self-esteem. Social Psychiatry & Psychiatric Epidemiology. Vol 26(1), pp. 39
Chafe, W. L. 1980. The development of consciousness in the production of narrative. The pear stories: cognitive, cultural, and linguistic aspects of narrative production. W. L. Chafe. Norwood, New Jersey, Ablex Publishing Company.
Cohen, Jacob. 1994. The Earth is Round (p<.05). American Psychologist 49: 997-1003
Cook, D.R. 1989. The Internalized Shame Scale. Stout, Wisconsin.
Durkheim, Emile. 1901. Sucide. Glencoe, Ill.: Free Press (1951).
Fearon, David. 1994. The Bond Threat Sequence: Discourse Evidence for the Interdependence of Shame and Social Relationships. Unpublished MA thesis, Dept of Sociology, UCSB.
Feynman, Richard. Surely You’re Joking, Mr. Feynman. New York: Norton.
Gottschalk, L., C. Winget, and G. Gleser. 1969. Manual of Instructions for Using the Gottschalk-Gleser Content Analysis Scales. Berkeley and Los Angeles: University of California Press.
Gottschalk, Louis. 1995. Content Analysis of Verbal Behavior. Hillsdale, NJ: Lawrence Erlbaum Associates.
Gumperz, J. J. 1982. Discourse strategies. Cambridge, Cambridge University Press.
Helmreich, R., & Stapp, J. (1974). Short forms of the Texas Social Behavior Inventory (TSBI), an objective measure of self-esteem. Bulletin of the Psychonomic Society, 4(5A), 473-475.
Heritage, J. 1984. Garfinkel and Ethnomethodology. New York, Polity Press.
Jackson, M. 1984. Self-Esteem and Meaning: A Life-Historical Investigation. Albany, N.Y.: SUNY Press.
Kaplan, Howard. 2001. Self-Esteem and Deviant Behavior. Ch. 17 in Owens, T.J., S. Stryker, and N. Goodman (Editors), Extending Self-Esteem Theory and Research. Cambridge: Cambridge University Press.
Kling, Kristen C; Hyde, Janet Shibley; Showers, Carolin J; Buswell, Brenda N. 1999. Gender differences in self-esteem: A meta-analysis. Psychological Bulletin. Vol 125(4), 470-500.
Lammers, H. B., & Becker, L. A. (1992). Distraction, choice and self-esteem effects on cognitive response facilitation. British Journal of Social Psychology(31), 189-200.
Leary, Mark and Roy Baumeister. 2000. The Nature and Function of Self-Esteem: Sociometer Theory. Pp. 1-62 in Advances in Experimental Social Psychology. Mark Zanna, (Editor). V. 32. San Diego: Academic Press.
Levinson, S. C. 1983. Pragmatics. Cambridge, Cambridge University Press.
Lewis, Helen B. 1971. Shame and Guilt in Neurosis. New York: International Universities Press.
Mruk, Christopher. 1999. Self Esteem: Research, Theory, and Practice. New York: Springer.
Owens, Timothy, and Adam King. 2001. Measuring Self-Esteem: Race, Ethnicity, and Gender Reconsidered. Pp. 56-84, in Owens, T.J., S. Stryker, and N. Goodman (Editors), Extending Self-Esteem Theory and Research. Cambridge: Cambridge University Press.
Retzinger, Suzanne. Violent Emotions: Shame and Rage in Marital Quarrels. Newbury Park, CA: Sage Publications.
Sarbin, Theodore, and James Mancuso. 1980. Elmsford, NY: Pergamon Press.
Scheff, Thomas. 1963. Decision Rules, Types of Error and Their Consequences in Medical Diagnosis. Behavior Science, 8 (April), pp. 97-107.
____________ 1990. Microsociology. Chicago: U. of Chicago Press.
_____________ 1994. Bloody Revenge. Boulder: Westview Press.
_____________ 1997. Emotions, the Social Bond, and Human Reality. Cambridge:
Cambridge U. Press.
Scheff, T., Retzinger, S. and Ryan, M. 1989. Crime, Violence and Self-Esteem. in A. Mecca, N. Smelser, and J. Vasconcellos (Editors), The Social Importance of Self-Esteem. Berkeley: U. of California Press.
Scheff, Thomas, and David Fearon. 2002. Shame and the Social Bond: Towards a Theory of Self-Esteem. (unpublished ms. 9,000 words).
Schneider, D.J., & Turkat, D. (1975). Self-presentation following success or failure: Defensive self-esteem models. Journal of Personality, 43, 127-135.
Schooler, C., and G. Oates. 2001. Self-Esteem and Work across the Life Course. Ch. 9 in in Owens, T.J., S. Stryker, and N. Goodman (Editors), Extending Self-Esteem Theory and Research. Cambridge: Cambridge University Press.
Selvin, Hanan. 1957. A critique of Tests of Significance in Survey Research. American Sociological Review. 22: 519-527.
Tangney, June, and Ronda Dearing. 2002. Shame and Guilt. New York: Guilford Press.
Twenge, Jean, and Jennifer Crocker. 2002. Race and Self-Esteem. Psychological Bulletin 128: 371-408.
Wells, L. E., and G. Marwell. 1976. Self-Esteem: Its Conceptualization and Measurement. Beverly Hills, Calif.: Sage.
Wells, L.E. 2001. Self-Esteem and Social Inequality. Ch. 14 in in Owens, T.J., S. Stryker, and N. Goodman (Editors), Extending Self-Esteem Theory and Research. Cambridge: Cambridge University Press.
Wilkerson, L. 1999. Statistical Methods in psychology journals: Guidelines and Explanations. American Psychologist 54, 594-604.
Wylie, R. C. 1979. The Self-Concept. Vol. 2, Theory and Research on Selected Topics. Rev. ed. Lincoln: University of Nebraska Press.
Ap1 2003 7708 words