The objective was to study the construct validity of the MOS SF-36 (Medical Outcome Study, short form questionnaire, version of 36 items) using structural equation modeling (SEM). Despite the widespread use of the questionnaire, several problems with the measurement model have been shown in the past, and it is highly disputable whether the computation of two component (summary) scales according to the published algorithm is a valid condensation. The SEM was conducted with the data of N= 2,874; the SF-36 was part of a questionnaire given prior to acupuncture treatment. The results indicated poor construct validity of the proposed structure of the first and second order factor solutions. First and foremost, contrary to the algorithm stated in the manual, the two component scales are not uncorrelated. Second, the overall fit of the original measurement model does not reflect the data structure very well. The current measurement model is discussed and starting points for revisions are provided