Scaling assumptions and validity of the English (UK) and Chinese (HK) short form 36 health survey (SF-36) were assessed in a community-based survey of 5,503 Chinese, Malays and Indians in Singapore using the international quality of life assessment project approach of item and scale level validation. Missing data for SF-36 items and scales occurred in less than 1.0% of subjects. Item level validation of both versions generally supported assumptions underlying Likert scoring and hypothesised item-order clustering. Item level factor analysis supported the eight-scale structure of the SF-36. In scale level validation, SF-36 scale scores showed wide variability and acceptable internal-consistency reliability (Cronbach's alpha > 0.70 for six English and seven Chinese scales), conformed to hypothesised patterns and generally varied according to hypotheses in subjects known to differ in quality of life. Scale level factor analysis of both versions yielded very similar patterns of factor correlation, comparable to that found in Japan, but differing from that seen in Western populations. Taken together, these results support the validity of the English (UK) and Chinese (HK) SF-36 versions in the multi-ethnic Asian socio-cultural context of Singapore.