|
Article Excerpt
Differential item functioning (DIF) in the Multidimensional Self Concept Scale (B. A. Bracken, 1992) was evaluated using 2 different methods to identify and describe DIF. Of 149 items from the Multidimensional Self Concept Scale, 42% exhibited gender DIF. Nonuniform, crossover DIF was evident in items throughout the instrument.
**********
The relationship between gender and adolescent self-concept has been discussed extensively in the literature, with a variety of conclusions. Some researchers have reported significant differences that have tended to reflect gender stereotypes (American Association of University Women Educational Foundation, 1995; Block & Robins, 1993; Hirsch & Dubois, 1991; Simmons & Blyth, 1987; Wigfield, Eccles, MacIver, Reuman, & Midgley, 1991; Wilgenbusch & Merrell, 1999), whereas others have reported no practical differences (Crain & Bracken, 1994; Wylie, 1979).
Several facets of the research may contribute to these contradictory findings. A tendency to use small, nonrepresentative samples (Crain & Bracken, 1994) may be one contributing factor. Another reason the results have not been consistent may be that a variety of self-concept measures has been used, each having a distinct view of how self-concept or self-esteem is defined and measured. Another possibility is that items in the self-concept instruments function differently for males and females, thus contributing to these gender differences. Because of the variety of reported results, further inquiry is needed to explore what factors might be contributing to the gender differences and similarities in adolescent self-concept.
THEORETICAL ISSUES
Underlying the concerns in the gender and self-concept discussion are several theoretical and measurement issues. First, the fundamental nature of global self-concept continues to be discussed by prominent researchers. Harter (1999) claimed that self-concept is both a cognitive and a social construct. Harter (1999) focused on self-representations or "characteristics of the self that are consciously acknowledged by the individual through language" (p. 3). Furthermore, she has discussed how the developmental aspects of cognition (i.e., processes of developing abstract thought during adolescence) influence the development of self-concept. Older children and adolescents tend to be able to discriminate between the real and the ideal self and the self across social contexts. For example, the social nature of self-concept is developed as children react to other group experiences, such as caregiving and schooling. Harter (1999) developed several measures of self-perception that she used in her research.
Others have described self-concept as "a set of attitudes and beliefs" (Coopersmith, 1981, p. 1). Cooley (1902) considered self-concept to be a social construct. He suggested that what we believe others think of our characteristics creates a sense of self, what he called the "looking-glass self." Cooley also elaborated on the relationship between a developmental perspective and the inclusion of others' opinions in creating one's self-concept. Mead (1925, 1934) also stressed the importance of significant others' opinions in forming the basis of the self.
In contrast, Bracken (1992) indicated that self concept is a behavioral construct. (Bracken did not hyphenate the term to differentiate his construct from others.) A child's self concept reflects his or her "evaluation of past behaviors and experience, and predicts the individual's future behaviors" (p. 3). The Multidimensional Self Concept Scale (MSCS; Bracken, 1992), an outcome of this theory, uses a behavioral perspective that defines self concept as a learned pattern of behavior. An examinee's MSCS score reflects his or her self-evaluations based on the individual's history of reinforcement, social interactions, achievements, and failures.
The second theoretical issue has been the multidimensional nature of self-concept, which is now well established (Bracken, 1992; Byrne, 2002; Harter, 1996; Shavelson, Hubner, & Stanton, 1976). The research literature has moved from considering and defining only global self-concept to assessing separate and distinct domains (e.g., physical, academic, social). To better understand the distinctions between global self-concept and specific domains, Harter (1990) defined global self-worth as one's overall perceptions of personal worth. Domain-specific self-concept tends to be related to specific competencies or adequacies, such as academic, social, or physical (Harter, 1999). Bracken (1992) included six domains in his measure of self concept. He suggested there is overlap and interplay among these domains, yet each can be considered an independent construct. These domains are incorporated into a hierarchical model, with generalized self concept at the center and the various domains surrounding the core.
GENDER AND SELF-CONCEPT
These separate self-concept domains, which tend to be defined somewhat differently by each theorist or researcher, have become the central focus in understanding and researching many areas related to self-concept, in addition to gender differences and similarities. Recent research has presented some interesting gender differences within these specific self-concept domains. The results of a meta-analysis indicated that adolescent male participants reported a significantly higher self-concept than the adolescent female participants in the following areas: global, mathematics, physical appearance, athletic psychomotor coordination, and emotional/affect. Female participants' scores were higher than male participants' scores in only one area: verbal (Wilgenbusch & Merrell, 1999). However, one of the limits of this study was the variety of theoretical constructs reflected in the measures of self-concept and self-esteem used in the studies analyzed. Of the 22 studies included in the meta-analysis, 9 of the measures were developed by the respective study authors. The remaining 13 studies in the meta-analysis used 6 different measures of self-concept. These theoretical differences most likely had some influence on the research outcomes and seem to limit comparison among the studies.
In contrast to this finding, Crain and Bracken (1994), using a large, nationally representative sample of 2,501 students in Grades 5-12 (ages 9-19), reported no significant self-concept score differences in the means of the total raw scores for boys and girls. On the domain subscales, boys' scores were significantly higher on only the Physical subscale, but the authors suggested that this difference had minor clinical significance because the raw score difference was only about one-third standard deviation. In a current summary of the gender and self-concept research literature, Eccles, Barber, Jozefowics, Malenchuk, and Vida (1999) concluded that there was more variation in self-concept within gender groups than between groups. The authors asserted, "It is not the case that females, in general, have less self-esteem than males. Some do, but many do not. And in some domains, females in general have more confidence in their abilities than males" (p. 77). Although the conclusions about the relationship between gender and self-concept have been formed by considering mean score differences and similarities between genders in self-concept, considering the possibility of differential item functioning (DIF) takes the discussion one step further in investigating if and how the construct of self-concept varies for boys and girls.
DIF
As mentioned previously, the self-concept differences noted in the current body of research may result from using instruments that contain items that function differently for boys and girls. DIF methods have been used to understand the measurement procedures and theory underlying a variety of psychological constructs, such as intelligence (Maller, 2001), academic achievement (Snetzler & Qualls, 2000), or perceived stress (Cole, 1999). An item functions differentially when examinees, who have the same level of the targeted trait but who belong to different subgroups, tend to respond differently to the item because of group membership (Camilli & Shepard, 1994).
DIF may occur for a number of reasons. The item may have language that is offensive to members of a particular group, or the item may have negative wording that is confusing to one group. In addition, some test items present the content in ways that prevent certain groups of examinees from demonstrating their knowledge or traits. The way an item is worded or structured may introduce a source of difficulty that is not relevant to the construct being measured. Thus, the item may be measuring two or more constructs of which only one is relevant to the purpose of the instrument. The extraneous construct may influence the responses of one group differently. However, how it influences the primary construct may not be explicit (Camilli & Shepard, 1994). For example, on a test of math achievement, a weakness in reading comprehension may prevent a group of students from accurately...
|
