Cross-Cultural Bias in Likert-Type Inventories: Perfect Matching Factor Structures and Still Biased?

The academic self-concept, measured by the Reading self-concept scale and the Math self-concept scale of the “Self-Description-Questionnaire I” (SDQ-I; Marsh, 1988 ), of Singaporean Chinese sixth-graders (600 boys and 600 girls) was compared to those of a sample of 1200 Australian students of the same age and gender composition drawn from the SDQ-I calibration sample. As the Singaporeans were fluent in English, the original English scale was used so as to avoid possible translation bias. Each scale consists of four interest items, two competence items, and two task-easiness items. Subjects answered all items on a five-point rating scale ranging from “false” to “true.” Although the factor structure of the Singaporean Chinese resembled closely those of their Australian counterparts, substantial cross-cultural bias emerged between interest items and competence/easiness items when treated as a single scale. This is because Singaporean students as compared to the Australians were more reluctant to agree to items with self-praise connotation. In addition, cross-cultural differences were found in the psychological meaning of the rating categories.