OBJECTIVES: SF-36 is one of the most popular and extensively used health outcome measures. A large amount of resources have been devoted to examine the psychometric properties of SF-36, such as scale reliability, validity and the use of exploratory factor analysis or principal component analysis to identify the two latent constructs. However, few studies have examined the factor structure of SF-36 using the right tool -confirmatory factor analysis. None of the available studies have examined the factorial validity of SF-36v2. The study here is to investigate the factorial structure (construct validity) of SF-36v2 in the US general population using confirmatory factor analysis through structural equation modelling (SEM). METHODS: A sample consisting of 3844 non-institutionalized adults was drawn from the National Health Measurement Survey in the US and interviewed over the telephone. SF-36v2 was administered along with other health outcome measures. The factorial structure of SF-36 developed in Keller’s paper (1997) was adopted as a point of departure for the other models. The model’s goodness of fit to the data was evaluated using a number of fit indices, including CFI, TLI, SRMR, RMSEA, S-Bx2 Statistic and the corrected CFI. All analyses were performed using Mplus 3. RESULTS: Keller’s model provided an adequate fit to the data. However, other models that allowed for error terms of items to covary improved the model fit dramatically. The most interesting finding was that the model that contained only the measurement part performed best. That is, a hierarchic structure did not improve the model fit to the data regardless of the composition of higher order factors. CONCLUSIONS:The results supported the 8 hypothetical factors structure as proposed by the SF-36 developers. However, there were some doubts regarding the higher hierarchic construct of SF-36v2. Consequently, the use of two summary scores should be taken with caution.