Appendix:
Measures of Self-Esteem

In measuring any psychological construct, we want to know that the instrument accurately reflects the theoretical construct on which it is based and truthfully assesses the individual's internal reality. The first issue can be addressed by the measure's "face validity"—the extent to which individuals familiar with the theoretical construct agree that the measure adequately represents the construct; its "internal validity"—the degree to which the items in the measure intercorrelate, indicating a degree of psychological unity; and its "convergent validity"—the extent to which the measure is associated with other measures of the same construct.

Three additional types of evidence are relevant to determining whether the measure accurately describes the individual. The first is test-retest reliability—the extent to which individuals achieve the same scores relative to others in the sample over a short period of time, typically a few days to a few weeks. High test-retest reliability is consistent with the interpretation that the measure reflects something relatively enduring about the individual, rather than a fleeting or momentary mood state. The other two types of evidence are concurrent and predictive validity—the extent to which scores on the measure correlate with some other relevant measures assessed either at the same time (concurrently) or over a longer period (predictively).

Information concerning the reliability and validity of the three measures of self-esteem used most frequently in the studies we review—the Rosenberg Self-Esteem Scale (1965), the Coopersmith Self-Esteem Inventory (1967), and the Self-Derogation Scale from Kaplan, Smith, and Pokorny (1979)—is included below. We provide information on other measures when describing the individual study.

Rosenberg Self-Esteem Scale

The Rosenberg Self-Esteem Scale consists of ten items (e.g., "On the whole, I am satisfied with myself"), with one of four responses possible for each item: respondents are asked to strongly agree, agree, disagree, or strongly disagree with each test item.

Three properties of the measure are worth noting. First, the scale is designed to measure the respondents' global self-esteem . The items do not specify particular areas of activity or qualities that individuals must

take into consideration when judging themselves. The scale attempts to gauge a respondent's basic attitude toward his or her own worth by allowing individuals to invoke their own frame of reference. Second, the scale is designed to capture the respondent's enduring, longstanding self-estimate. The emphasis is not on one's immediate or momentary self-perception; rather, the scale stresses the more permanent, more stable components of the self-image. Finally, a high score on this scale does not mean that the respondent stands in awe of himself or herself, nor does it mean that a respondent expects such awe or deference from others. Rather, it reflects the feeling that one is "good enough"—a person of worth, who merits self-respect.

The psychometric properties of this scale have been summarized by Wylie (1974). She points out that the Rosenberg Scale's face validity appears to be good and that its convergence with other measures of self-esteem is acceptably high, ranging from .67 to .83. A two-week test-retest reliability coefficient is similarly high—.85 for twenty-eight college-age respondents. In addition, the scale appears to have the ability to predict and concur with behaviors, attitudes, and experiences to which self-esteem is theoretically expected to be related (e.g., depressive affect, anxiety and psychosomatic symptoms, interpersonal insecurity, participation in activities, leadership, and parental disinterest).

Regarding the scale's ability to provide an accurate assessment of the individual's internal reality, there remains, as there does for all self-concept measures, the possibility that respondents may distort reality in order to provide socially desirable answers. Researchers hope that if they are able to establish rapport and guarantee anonymity, respondents will answer truthfully.

Coopersmith Self-Esteem Inventory

The Coopersmith Self-Esteem Inventory (SEI) for children is a fifty-item questionnaire intended to measure the evaluation that children from ages eight to fifteen make and customarily maintain with regard to themselves. The questionnaire presents respondents with generally favorable or generally unfavorable statements about the self (e.g., "I am pretty sure of myself," "I'm easy to like"), which they designate as "like me" or "unlike me." Assuming that self-efficacy may vary across different areas of experience, Coopersmith includes questions from five different domains: peers, family, self, school, and general social activities.

The scores from these subscales are combined for a general self-esteem score. The scale is accompanied by an eight-item lie scale to assess defensiveness.

Coopersmith developed the measure with the assistance of several "self-esteem experts," and all the items in the final scale were agreed upon by five psychologists, supporting the scale's face validity. Coopersmith, like many self-esteem researchers, assumes that one's global self-esteem score consists of the sum of scores in five separate areas, as described above. Whether or not this assumption is valid is not known, for no item analyses or internal factor analyses have been performed. Wylie (1974) reports low to moderate convergence, ranging from .17 to .40, between scores on the Coopersmith SEI and other measures of self-esteem, as well as similarly moderate correlations with measures to which self-esteem is theoretically expected to be associated (e.g., the Iowa Achievement Tests, r = .36; and a sociometric rating of popularity, r = .29). Test-retest reliability is high: .88 over a five-week interval for fifth graders (Coopersmith 1967).

Coopersmith attempts to control for acquiescent responses by including an equal number of favorably and unfavorably worded items. Again, it is hoped that guaranteeing anonymity and establishing rapport with the respondents will cause them to answer truthfully rather than providing socially desirable answers.

Self-Derogation Scale

Kaplan's self-derogation scale consists of the seven items that make up the first factor derived from a factor analysis of the Rosenberg Scale (Kaplan and Pokorny 1969). The scale indexes a lack of pride and self-respect and feelings of personal failure and worthlessness. Thus a low score indicates high self-esteem. Because the scale is made up of items from the Rosenberg Self-Esteem Scale, we may presume it is equally valid on its face.

The inter-item correlations for the seven items making up the scale, as reported by Kaplan (1980), are not high but are statistically significant. The scale's internal consistency is estimated at .79 (Cronbach's alpha) (Robbins, Kaplan, and Martin 1985). Kaplan does not report the scale's convergence with other measures of self-esteem, but it may be presumed that its convergent validity does not differ appreciably from the Rosenberg Self-Esteem Scale, on which it is based.

Unlike Coopersmith, Kaplan does not expect high long-term test-retest reliability scores, especially for children with low self-esteem. Accordingly, test-retest correlations ranged from .55 (over one year) to .40 (over a two-year interval) during adolescence. In addition, Kaplan and Pokorny (1969) report that high scores on the self-derogation scale are associated with several measures of psychosocial adjustment, including self-reports of psychological symptoms, scores on a depressive affect scale, and use of psychiatric services during the preceding year.