Construct validation of the Bem Sex Role Inventory in Taiwan

Sex Roles: A Journal of Research,  Dec, 2006  by T.K. Peng

• E-mail
• Print
• Link

Since the formulation of the Bem Sex Role Inventory (BSRI) in the 1970s (Bem, 1974), researchers in North America and elsewhere have continued to examine its psychometric properties under various contexts. Although the BSRI is the most frequently tested gender role instrument and one of the most investigated psychometric constructs in general (Lenney, 1991), its application to cultural settings other than North America is far from clear. Moreover, no study in the Western literature so far has documented the reliability and validity of the BSRI with research participants in Taiwan. Thus, a link with regard to the BSRI's cross-cultural applicability is therefore missing from the literature.

The present study is an attempt to fill this gap by empirically testing the psychometric properties of the instrument with several occupational groups in Taiwan. Unlike mainland China and Hong Kong, which have been strongly influenced by communism and Western culture, respectively, Taiwan remains to be a Confucian society that preserves Chinese values relatively well (Redding, 1990). Therefore, it is interesting to explore how BSRI applies in this society.

Validation of the BSRI in North America and Chinese Contexts

At the outset, Bem (1974) designed the BSRI as an instrument to measure masculinity (M, thereinafter) and femininity (F, thereinafter) with separate scales, rather than one unidimensional scale that considers M and F to be polar opposites. Based on gender schema theory (Bem, 1981), the perception of M and F is a culture-related process through which individuals organize information about self and others. Sandra Bem stressed that the BSRI measures the level of desirability of characteristics for men or for women rather than how much research participants actually show these characteristics themselves. The original BSRI consists of 60 items: 20 stereotypically masculine, 20 stereotypically feminine, and 20 neutral filler items. Moreover, the BSRI manual (Bem, 1981) reports original tests in the US that resulted in internal consistencies that range from 0.80 to 0.86. The test-retest reliabilities of M were 0.94 and 0.76, whereas those of F were 0.90 and 0.89, for women and men, respectively. Moreover, the self-rated gender role scores can be subjected a median split to categorize the respondents as M, F, androgynous, or undifferentiated. These early studies also revealed that, as expected, men scored higher on the M dimension whereas women scored higher on the F dimension.

Most of the studies that assessed the reliability and validity of the BSRI have been conducted by scholars in the United States with samples from North America; those based on other cultural contexts are sporadically few (e.g., OcKany & Lajunen, 2005; Sugihara & Katsurada, 1999). In the interest of brevity, the review below first highlights the findings from studies with a North American focus, and then summarizes the results from Chinese samples.

North America Over the years, there have been quite a few studies of the psychometric properties of the BSRI in North America. Pedhazur and Tetenbaum (1979) critically challenged Bem's inventory in terms of theoretical basis, item selection, and dimensionality. For example, they questioned that some of the gender traits were not necessarily as desirable (i.e., positive) as the original framework had claimed. In response to such criticisms, Bem (1981) dropped some of the items and developed the short form of BSRI, which consists of 30 items and appears to be more psychologically sound than the longer form (Campbell, Gillaspy, & Thompson, 1997). Choi and Fuqua (2003) located 23 BSRI validation studies published during 1974-1992 and reviewed the factor structures of the instrument revealed in these studies. They found that most studies yielded one clean F factor, and only a few reported more than one F factor, whereas the M factor appeared to be more complex because the majority of researchers reported two to three factors from the M items. Because this pattern in the structure was quite consistent, Choi and Fugua suspected that it had to do with the relative homogeneity of the samples in that three-fourths of the samples in the 23 studies were college students.

Partly due to the changes in the roles of women and men since the 1970s in the United States, several validity studies published in recent years concern the current theoretical meaningfulness of the BSRI in today's environment. Hoffman and Borders (2001) questioned the theoretical rationale, classifications, and scoring methods proposed by Bem (1981). A comparison of the median scores of M and F between their study and those reported in Bem indicates a similar range. The correlation between the long and short forms for the Hoffman and Borders sample was 0.92, which suggests that the two forms functioned similarly. However, different scoring methods (i.e., median-split, the most popular, vs. hybrid) resulted in different gender role classifications. When the forms (i.e., long and short) were considered, the difference in classification became even more complicated. It was also found that the gender role perceptions of contemporary American college students deviated markedly from those of their counterparts in earlier days. Hoffman and Borders concluded that gender schema theory (Bem, 1981) may be less relevant now than before. In a separate attempt, Holt and Ellis (1998) partially replicated Bem's (1974) procedure to validate the original 60-item BSRI. Their results reflect minor discrepancies from the original. The reliability coefficients for the two scales were both over 0.90. Their assessment shows that the BSRI is still valid, but traditional M and F perceptions may be weakening. Although the review here is far from comprehensive, it is reasonable to conclude that the validity of the BSRI in the US continues to be controversial decades after its creation.