Interpreting scale items: using items on the Bem Sex Role Inventory to explore respondents' meaning construction

Chu, Ping-Fang Martha
Journal Title
Journal ISSN
Volume Title

Quantitative researchers often infer the validity of a construct through some quantitative (statistical) manipulation of the numerical responses to scale or questionnaire items. The participants' meaning construction of these items, especially with affective and personality scales, are usually unexplored or assumed equivalent. The current study investigates the defensibility of such a position based on the socio-constructive view on meaning construction. The current study explored how respondents interpreted the items from the Bem Sex Role Inventory (BSRI). The instrument was chosen for its popularity, its empirically based development, and for the culturally significant issue the scale addresses. Bem also proposed a gender schema theory to account for sex related cognitive processing. Forty participants, equally divided into male and female, undergraduate and working adult groups, completed self-ratings on the Masculinity and Femininity items from the BSRI. They performed "think aloud" on half of the items while rating themselves. They also completed a follow-up interview providing more in-depth interpretations of scale items and their conceptualization of masculinity and femininity. The "think aloud" and the interviews were audiotape-recorded, transcribed, and analyzed. The results showed that the respondents constructed various meanings for each item. There were commonly shared meanings, less shared meanings, rare, and unique meanings. There were also misinterpretations. Participants conceptualized "being masculine" and "being feminine" in many dimensions, including physical features and personal appearance. The current study identified sex differences in meaning constructions, including the contexts in which meanings were situated. The current study also found substantial differences between the student and the adult participants' meaning constructions. The limitations of the current study and its implications for test construction are discussed.