As a foundational instrument in the measurement of gender, the study investigated the psychometric properties of the item scores on the 30-item Bem Sex Roles Inventory (BSRI) to address concerns raised in previous studies and to determine possible differential item functioning (DIF) by applying a Rasch measurement model on a multicultural South African sample. A quantitative cross-sectional research design with 55.3% women and 44.4% men was used. Overall, we found support for the reliability and validity of BSRI item scores, consistent with findings in North American and European contexts. The results supported the use of a 6-point response format as opposed to the original 7-point response format. The items "aggressive" (masculinity) and "loves children" (femininity) were removed. The Neutral dimension fit the data well. No DIF for the groupings of man/woman, race, and language (the latter 2 as measures of culture) were found. This study makes contributions by applying a Rasch measurement model to the BSRI, scores which have previously only been validated through classical test theory and often excludes the validation of the neutral dimension scores.