OBJECTIVES: To investigate the comparability of the English, Chinese and Malay versions of the 5 EQ-5D-5L dimension scales in Singapore. METHODS: A cross-sectional survey was conducted among visitors to a public primary care institution in Singapore. Participants were interviewed face-to-face in a language they preferred to measure the severity of the EQ-5D-5L response labels (‘no’, ‘slight’, ‘moderate’, ‘severe’, and ‘unable to (do)/extreme’ for each dimension scale) they perceived, using a 0-100 numeric rating scale. Participants were also asked to describe 25 dimension-specific health scenarios (5 for each dimension), each using one of the EQ-5D-5L response labels. Differences in label ratings and selection between participants speaking different languages were assessed using regression analysis. The individual-level rank order of label ratings was examined to assess violation of scale ordinality. RESULTS: : Severity of the EQ-5D-5L response labels perceived by Chinese-speaking (n=256) and English-speaking participants (n=257) was similar. Malay-speaking participants (n=230) perceived the ‘slight’ labels as more severe (adjusted differences in severity score: 10.5 to 14.5) but the ‘unable to (do)/extreme’ labels as less severe than English-speaking participants (adjusted differences in severity score: -13.3 to -11.0) (p < 0.001 for all). While no major violation of ordinality was observed among English- and Chinese-speaking participants,  a noticeable proportion of Malay-speaking participants rated ‘slight’ as more severe than ‘moderate’ (up to 29.6%)  and  ‘severe’ as more severe than ‘unable to (do)/extreme’ (up to 33.9%). In label selection exercise, Chinese- speaking participants tended to use less severe labels than English-speaking participants; Malay-speaking participants tended to use ‘no’ and ‘unable to (do)/extreme’. CONCLUSIONS: : Differences in perception and usage of response labels exist among Singaporeans using different language versions of the EQ-5D-5L dimension scales. Future studies are needed to investigate whether the differences affect the cross-cultural measurement equivalence of the instrument and, if so, how to alleviate the differences.