Abstract

This study addressed two problems that have been repeatedly observed in the previous studies using Korean Sex Role Inventory(KSRI). The first problem was that far too many male respondents had been classified as the androgynous type, and disproportionately large number of female respondents as the undifferentiated type by the KSRI. The second problem was that the same respondent could be classified into different sex role categories because sample medians were used as the cut-off points for classification. A previous study which examined the psychometric properties of the KSRI items failed to discover any psychometric defects among the items. Thus, study 1 was designed to test the hypothesis that the unexpected distribution of sex role categories is due to resoonse bias. That is, males respond to the items by comparing themselves with other males in general, and females respond by comparing themselves with other females in general. Four versions of KSRI were constructed by changing the instruction requesting the respondents to rate the items by comparing themselves with other males, females, both or each. The hypothesis was clearly confirmed. It was concluded that the disproportionate numbers of male respondents in the androgynous category and female respondents in the undifferentiated category were the result of response bias with the respondents rating themselves by applying stereotypical standards with respect to their own biological gender. Study 2 based on the data collected in study 1, attempted to derive adjustment coefficients that can be added ito the raw scores of KSRI in order to estimate the scores that would be obtained if the respondents have had applied gender-free judgmental standards. When the adjusted scores were used to classify the respondents into the sex role categories, the unexpected distribution of sex role categories no longer occurred. Finally, study 3 determined permanent cut-off scores that should be used irrespective of particular samples. For this determination, all previous data on KSRI were combined(n= 1675) and subjected to Item Response Theory analysis. Cut-off points of the femininity and masculinity subscales were determined at the scores that pro­ vide the highest test information. Since the higest test information implies the strongest discriminating power, these cut-off scores are theoretically considered to guarantee the lowest rates of misclassification. With the adjustment coefficients and the sample-free cut-off points the previous problems associated with KSRI will no longer occur. Follow-up studies need to examine the validities of the coefficients and the cut-off points. Implications for Bern's original Sex Role Inventory were discussed.